X-Git-Url: https://git.exim.org/exim.git/blobdiff_plain/74f1a42304ce056cf979d22fb970faae161e3ab2..a85c067ba6c6940512cf57ec213277a370d87e70:/src/src/malware.c diff --git a/src/src/malware.c b/src/src/malware.c index 49456de4e..96514e276 100644 --- a/src/src/malware.c +++ b/src/src/malware.c @@ -2,9 +2,11 @@ * Exim - an Internet mail transport agent * *************************************************/ -/* Copyright (c) Tom Kistner 2003 - 2015 +/* + * Copyright (c) The Exim Maintainers 2015 - 2022 + * Copyright (c) Tom Kistner 2003 - 2015 * License: GPL - * Copyright (c) The Exim Maintainers 2015 - 2018 + * SPDX-License-Identifier: GPL-2.0-only */ /* Code for calling virus (malware) scanners. Called from acl.c. */ @@ -49,6 +51,7 @@ typedef enum { #ifndef DISABLE_MAL_CMDLINE M_CMDL, #endif + M_DUMMY } scanner_t; typedef enum {MC_NONE, MC_TCP, MC_UNIX, MC_STRM} contype_t; static struct scan @@ -106,16 +109,16 @@ static struct scan void features_malware(void) { -const struct scan * sc; const uschar * s; uschar * t; -uschar buf[64]; +uschar buf[EXIM_DRIVERNAME_MAX]; spf(buf, sizeof(buf), US"_HAVE_MALWARE_"); -for (sc = m_scans; sc->scancode != -1; sc++) +for (const struct scan * sc = m_scans; sc->scancode != -1; sc++) { - for(s = sc->name, t = buf+14; *s; s++) if (*s != '-') *t++ = toupper(*s); + for (s = sc->name, t = buf+14; *s; s++) if (*s != '-') + *t++ = toupper(*s); *t = '\0'; builtin_macro_create(buf); } @@ -128,8 +131,7 @@ for (sc = m_scans; sc->scancode != -1; sc++) #define MALWARE_TIMEOUT 120 /* default timeout, seconds */ static const uschar * malware_regex_default = US ".+"; -static const pcre * malware_default_re = NULL; - +static const pcre2_code * malware_default_re = NULL; #ifndef DISABLE_MAL_CLAM @@ -156,46 +158,47 @@ typedef struct clamd_address { # define DERR_BAD_CALL (1<<15) /* wrong command */ static const uschar * drweb_re_str = US "infected\\swith\\s*(.+?)$"; -static const pcre * drweb_re = NULL; +static const pcre2_code * drweb_re = NULL; #endif #ifndef DISABLE_MAL_FSECURE static const uschar * fsec_re_str = US "\\S{0,5}INFECTED\\t[^\\t]*\\t([^\\t]+)\\t\\S*$"; -static const pcre * fsec_re = NULL; +static const pcre2_code * fsec_re = NULL; #endif #ifndef DISABLE_MAL_KAV static const uschar * kav_re_sus_str = US "suspicion:\\s*(.+?)\\s*$"; static const uschar * kav_re_inf_str = US "infected:\\s*(.+?)\\s*$"; -static const pcre * kav_re_sus = NULL; -static const pcre * kav_re_inf = NULL; +static const pcre2_code * kav_re_sus = NULL; +static const pcre2_code * kav_re_inf = NULL; #endif #ifndef DISABLE_MAL_AVAST static const uschar * ava_re_clean_str = US "(?!\\\\)\\t\\[\\+\\]"; static const uschar * ava_re_virus_str = US "(?!\\\\)\\t\\[L\\]\\d+\\.0\\t0\\s(.*)"; static const uschar * ava_re_error_str = US "(?!\\\\)\\t\\[E\\]\\d+\\.0\\tError\\s\\d+\\s(.*)"; -static const pcre * ava_re_clean = NULL; -static const pcre * ava_re_virus = NULL; -static const pcre * ava_re_error = NULL; +static const pcre2_code * ava_re_clean = NULL; +static const pcre2_code * ava_re_virus = NULL; +static const pcre2_code * ava_re_error = NULL; #endif #ifndef DISABLE_MAL_FFROT6D static const uschar * fprot6d_re_error_str = US "^\\d+\\s<(.+?)>$"; static const uschar * fprot6d_re_virus_str = US "^\\d+\\s\\s+.+$"; -static const pcre * fprot6d_re_error = NULL; -static const pcre * fprot6d_re_virus = NULL; +static const pcre2_code * fprot6d_re_error = NULL; +static const pcre2_code * fprot6d_re_virus = NULL; #endif /******************************************************************************/ +#ifndef DISABLE_MAL_KAV /* Routine to check whether a system is big- or little-endian. Ripped from http://www.faqs.org/faqs/graphics/fileformats-faq/part4/section-7.html Needed for proper kavdaemon implementation. Sigh. */ -#define BIG_MY_ENDIAN 0 -#define LITTLE_MY_ENDIAN 1 +# define BIG_MY_ENDIAN 0 +# define LITTLE_MY_ENDIAN 1 static int test_byte_order(void); static inline int test_byte_order() @@ -204,6 +207,7 @@ test_byte_order() char *byte = CS &word; return(byte[0] ? LITTLE_MY_ENDIAN : BIG_MY_ENDIAN); } +#endif BOOL malware_ok = FALSE; @@ -217,6 +221,7 @@ extern uschar spooled_message_id[MESSAGE_ID_LENGTH+1]; /* Some (currently avast only) use backslash escaped whitespace, this function undoes these escapes */ +#ifndef DISABLE_MAL_AVAST static inline void unescape(uschar *p) { @@ -225,6 +230,7 @@ for (; *p; ++p) if (*p == '\\' && (isspace(p[1]) || p[1] == '\\')) for (p0 = p; *p0; ++p0) *p0 = p0[1]; } +#endif /* --- malware_*_defer --- */ static inline int @@ -247,28 +253,15 @@ m_panic_defer(struct scan * scanent, const uschar * hostport, return malware_panic_defer(string_sprintf("%s %s : %s", scanent->name, hostport ? hostport : CUS"", str)); } -static inline int -m_log_defer(struct scan * scanent, const uschar * hostport, - const uschar * str) -{ -return malware_log_defer(string_sprintf("%s %s : %s", - scanent->name, hostport ? hostport : CUS"", str)); -} /* --- m_*_defer_3 */ static inline int m_panic_defer_3(struct scan * scanent, const uschar * hostport, const uschar * str, int fd_to_close) { +DEBUG(D_acl) debug_print_socket(fd_to_close); (void) close(fd_to_close); return m_panic_defer(scanent, hostport, str); } -static inline int -m_log_defer_3(struct scan * scanent, const uschar * hostport, - const uschar * str, int fd_to_close) -{ -(void) close(fd_to_close); -return m_log_defer(scanent, hostport, str); -} /*************************************************/ @@ -280,8 +273,15 @@ static inline int m_tcpsocket(const uschar * hostname, unsigned int port, host_item * host, uschar ** errstr, const blob * fastopen_blob) { -return ip_connectedsocket(SOCK_STREAM, hostname, port, port, 5, +int fd = ip_connectedsocket(SOCK_STREAM, hostname, port, port, 5, host, errstr, fastopen_blob); +#ifdef EXIM_TFO_FREEBSD +/* Under some fault conditions, FreeBSD 12.2 seen to send a (non-TFO) SYN +and, getting no response, wait for a long time. Impose a 5s max. */ +if (fd >= 0) + (void) poll_one_fd(fd, POLLOUT, 5 * 1000); +#endif +return fd; } #endif @@ -299,38 +299,36 @@ if (send(sock, buf, cnt, 0) < 0) return sock; } -static const pcre * -m_pcre_compile(const uschar * re, uschar ** errstr) +static const pcre2_code * +m_pcre_compile(const uschar * re, BOOL cacheable, uschar ** errstr) { -const uschar * rerror; -int roffset; -const pcre * cre; - -cre = pcre_compile(CS re, PCRE_COPT, (const char **)&rerror, &roffset, NULL); -if (!cre) - *errstr= string_sprintf("regular expression error in '%s': %s at offset %d", - re, rerror, roffset); -return cre; +return regex_compile(re, cacheable ? MCS_CACHEABLE : MCS_NOFLAGS, errstr, + pcre_gen_cmp_ctx); } uschar * -m_pcre_exec(const pcre * cre, uschar * text) +m_pcre_exec(const pcre2_code * cre, uschar * text) { -int ovector[10*3]; -int i = pcre_exec(cre, NULL, CS text, Ustrlen(text), 0, 0, - ovector, nelem(ovector)); +pcre2_match_data * md = pcre2_match_data_create(2, pcre_gen_ctx); +int i = pcre2_match(cre, text, PCRE2_ZERO_TERMINATED, 0, 0, md, pcre_gen_mtc_ctx); uschar * substr = NULL; + if (i >= 2) /* Got it */ - pcre_get_substring(CS text, ovector, i, 1, (const char **) &substr); + { + PCRE2_SIZE * ovec = pcre2_get_ovector_pointer(md); + int len = ovec[3] - ovec[2]; + substr = string_copyn(text + ovec[2], len); + } +/* pcre2_match_data_free(md); gen ctx needs no free */ return substr; } -static const pcre * +static const pcre2_code * m_pcre_nextinlist(const uschar ** list, int * sep, - char * listerr, uschar ** errstr) + BOOL cacheable, char * listerr, uschar ** errstr) { const uschar * list_ele; -const pcre * cre = NULL; +const pcre2_code * cre = NULL; if (!(list_ele = string_nextinlist(list, sep, NULL, 0))) *errstr = US listerr; @@ -338,11 +336,12 @@ else { DEBUG(D_acl) debug_printf_indent("%15s%10s'%s'\n", "", "RE: ", string_printing(list_ele)); - cre = m_pcre_compile(CUS list_ele, errstr); + cre = m_pcre_compile(CUS list_ele, cacheable, errstr); } return cre; } + /* Simple though inefficient wrapper for reading a line. Drop CRs and the trailing newline. Can return early on buffer full. Null-terminate. @@ -353,13 +352,13 @@ return cre; -2 on timeout or error */ static int -recv_line(int fd, uschar * buffer, int bsize, int tmo) +recv_line(int fd, uschar * buffer, int bsize, time_t tmo) { uschar * p = buffer; ssize_t rcv; BOOL ok = FALSE; -if (!fd_ready(fd, tmo-time(NULL))) +if (!fd_ready(fd, tmo)) return -2; /*XXX tmo handling assumes we always get a whole line */ @@ -374,8 +373,12 @@ while ((rcv = read(fd, p, 1)) > 0) } if (!ok) { - DEBUG(D_acl) debug_printf_indent("Malware scan: read %s (%s)\n", + DEBUG(D_acl) + { + debug_printf_indent("Malware scan: read %s (%s)\n", rcv==0 ? "EOF" : "error", strerror(errno)); + debug_print_socket(fd); + } return rcv==0 ? -1 : -2; } *p = '\0'; @@ -385,13 +388,15 @@ return p - buffer; } /* return TRUE iff size as requested */ +#ifndef DISABLE_MAL_DRWEB static BOOL -recv_len(int sock, void * buf, int size, int tmo) +recv_len(int sock, void * buf, int size, time_t tmo) { -return fd_ready(sock, tmo-time(NULL)) +return fd_ready(sock, tmo) ? recv(sock, buf, size, 0) == size : FALSE; } +#endif @@ -434,7 +439,7 @@ for (;;) } static inline int -mksd_read_lines (int sock, uschar *av_buffer, int av_buffer_size, int tmo) +mksd_read_lines (int sock, uschar *av_buffer, int av_buffer_size, time_t tmo) { client_conn_ctx cctx = {.sock = sock}; int offset = 0; @@ -442,7 +447,7 @@ int i; do { - i = ip_recv(&cctx, av_buffer+offset, av_buffer_size-offset, tmo-time(NULL)); + i = ip_recv(&cctx, av_buffer+offset, av_buffer_size-offset, tmo); if (i <= 0) { (void) malware_panic_defer(US"unable to read from mksd UNIX socket (/var/run/mksd/socket)"); @@ -501,7 +506,7 @@ switch (*line) static int mksd_scan_packed(struct scan * scanent, int sock, const uschar * scan_filename, - int tmo) + time_t tmo) { struct iovec iov[3]; const char *cmd = "MSQ\n"; @@ -557,6 +562,7 @@ is via malware(), or there's malware_in_file() used for testing/debugging. Arguments: malware_re match condition for "malware=" + cacheable the RE did not use any dynamic elements during expansion scan_filename the file holding the email to be scanned, if we're faking this up for the -bmalware test, else NULL timeout if nonzero, non-default timeoutl @@ -565,15 +571,16 @@ Returns: Exim message processing code (OK, FAIL, DEFER, ...) where true means malware was found (condition applies) */ static int -malware_internal(const uschar * malware_re, const uschar * scan_filename, - int timeout) +malware_internal(const uschar * malware_re, BOOL cacheable, + const uschar * scan_filename, int timeout) { int sep = 0; const uschar *av_scanner_work = av_scanner; +BOOL av_scanner_textonly; uschar *scanner_name; unsigned long mbox_size; FILE *mbox_file; -const pcre *re; +const pcre2_code *re; uschar * errstr; struct scan * scanent; const uschar * scanner_options; @@ -596,30 +603,30 @@ the name), so we can close it right away. Get the directory too. */ eml_dir = string_copyn(eml_filename, Ustrrchr(eml_filename, '/') - eml_filename); /* parse 1st option */ -if (strcmpic(malware_re, US"false") == 0 || Ustrcmp(malware_re,"0") == 0) +if (strcmpic(malware_re, US"false") == 0 || Ustrcmp(malware_re, "0") == 0) return FAIL; /* explicitly no matching */ /* special cases (match anything except empty) */ -if ( strcmpic(malware_re,US"true") == 0 - || Ustrcmp(malware_re,"*") == 0 - || Ustrcmp(malware_re,"1") == 0 +if ( strcmpic(malware_re, US"true") == 0 + || Ustrcmp(malware_re, "*") == 0 + || Ustrcmp(malware_re, "1") == 0 ) { if ( !malware_default_re - && !(malware_default_re = m_pcre_compile(malware_regex_default, &errstr))) + && !(malware_default_re = m_pcre_compile(malware_regex_default, FALSE, &errstr))) return malware_panic_defer(errstr); malware_re = malware_regex_default; re = malware_default_re; } /* compile the regex, see if it works */ -else if (!(re = m_pcre_compile(malware_re, &errstr))) +else if (!(re = m_pcre_compile(malware_re, cacheable, &errstr))) return malware_panic_defer(errstr); /* if av_scanner starts with a dollar, expand it first */ if (*av_scanner == '$') { - if (!(av_scanner_work = expand_string(av_scanner))) + if (!(av_scanner_work = expand_string_2(av_scanner, &av_scanner_textonly))) return malware_panic_defer( string_sprintf("av_scanner starts with $, but expansion failed: %s", expand_string_message)); @@ -630,6 +637,8 @@ if (*av_scanner == '$') malware_name = NULL; malware_ok = FALSE; } +else + av_scanner_textonly = TRUE; /* Do not scan twice (unless av_scanner is dynamic). */ if (!malware_ok) @@ -659,11 +668,11 @@ if (!malware_ok) switch(scanent->conn) { case MC_TCP: - malware_daemon_ctx.sock = ip_tcpsocket(scanner_options, &errstr, 5); break; + malware_daemon_ctx.sock = ip_tcpsocket(scanner_options, &errstr, 5, NULL); break; case MC_UNIX: malware_daemon_ctx.sock = ip_unixsocket(scanner_options, &errstr); break; case MC_STRM: - malware_daemon_ctx.sock = ip_streamsocket(scanner_options, &errstr, 5); break; + malware_daemon_ctx.sock = ip_streamsocket(scanner_options, &errstr, 5, NULL); break; default: /* compiler quietening */ break; } @@ -734,13 +743,11 @@ if (!malware_ok) case M_FPROT6D: /* "f-prot6d" scanner type ----------------------------------- */ { int bread; - uschar * e; - uschar * linebuffer; - uschar * scanrequest; + uschar * e, * linebuffer, * scanrequest; uschar av_buffer[1024]; - if ((!fprot6d_re_virus && !(fprot6d_re_virus = m_pcre_compile(fprot6d_re_virus_str, &errstr))) - || (!fprot6d_re_error && !(fprot6d_re_error = m_pcre_compile(fprot6d_re_error_str, &errstr)))) + if ((!fprot6d_re_virus && !(fprot6d_re_virus = m_pcre_compile(fprot6d_re_virus_str, FALSE, &errstr))) + || (!fprot6d_re_error && !(fprot6d_re_error = m_pcre_compile(fprot6d_re_error_str, FALSE, &errstr)))) return malware_panic_defer(errstr); scanrequest = string_sprintf("SCAN FILE %s\n", eml_filename); @@ -750,7 +757,7 @@ if (!malware_ok) if (m_sock_send(malware_daemon_ctx.sock, scanrequest, Ustrlen(scanrequest), &errstr) < 0) return m_panic_defer(scanent, CUS callout_address, errstr); - bread = ip_recv(&malware_daemon_ctx, av_buffer, sizeof(av_buffer), tmo-time(NULL)); + bread = ip_recv(&malware_daemon_ctx, av_buffer, sizeof(av_buffer), tmo); if (bread <= 0) return m_panic_defer_3(scanent, CUS callout_address, @@ -796,7 +803,7 @@ if (!malware_ok) if (*scanner_options != '/') { /* calc file size */ - if ((drweb_fd = open(CCS eml_filename, O_RDONLY)) == -1) + if ((drweb_fd = exim_open2(CCS eml_filename, O_RDONLY)) == -1) return m_panic_defer_3(scanent, NULL, string_sprintf("can't open spool file %s: %s", eml_filename, strerror(errno)), @@ -840,7 +847,7 @@ badseek: err = errno; malware_daemon_ctx.sock); } - if (!(drweb_fbuf = US malloc(fsize_uint))) + if (!(drweb_fbuf = store_malloc(fsize_uint))) { (void)close(drweb_fd); return m_panic_defer_3(scanent, NULL, @@ -853,7 +860,7 @@ badseek: err = errno; { int err = errno; (void)close(drweb_fd); - free(drweb_fbuf); + store_free(drweb_fbuf); return m_panic_defer_3(scanent, NULL, string_sprintf("can't read spool file %s: %s", eml_filename, strerror(err)), @@ -864,11 +871,12 @@ badseek: err = errno; /* send file body to socket */ if (send(malware_daemon_ctx.sock, drweb_fbuf, fsize, 0) < 0) { - free(drweb_fbuf); + store_free(drweb_fbuf); return m_panic_defer_3(scanent, CUS callout_address, string_sprintf( "unable to send file body to socket (%s)", scanner_options), malware_daemon_ctx.sock); } + store_free(drweb_fbuf); } else { @@ -902,7 +910,6 @@ badseek: err = errno; /* "virus(es) found" if virus number is > 0 */ if (drweb_vnum) { - int i; gstring * g = NULL; /* setup default virus name */ @@ -910,19 +917,21 @@ badseek: err = errno; /* set up match regex */ if (!drweb_re) - drweb_re = m_pcre_compile(drweb_re_str, &errstr); + drweb_re = m_pcre_compile(drweb_re_str, FALSE, &errstr); /* read and concatenate virus names into one string */ - for (i = 0; i < drweb_vnum; i++) + for (int i = 0; i < drweb_vnum; i++) { - int ovector[10*3]; + pcre2_match_data * md = pcre2_match_data_create(2, pcre_gen_ctx); /* read the size of report */ if (!recv_len(malware_daemon_ctx.sock, &drweb_slen, sizeof(drweb_slen), tmo)) return m_panic_defer_3(scanent, CUS callout_address, US"cannot read report size", malware_daemon_ctx.sock); drweb_slen = ntohl(drweb_slen); - tmpbuf = store_get(drweb_slen); + + /* assume tainted, since it is external input */ + tmpbuf = store_get(drweb_slen, GET_TAINTED); /* read report body */ if (!recv_len(malware_daemon_ctx.sock, tmpbuf, drweb_slen, tmo)) @@ -931,23 +940,22 @@ badseek: err = errno; tmpbuf[drweb_slen] = '\0'; /* try matcher on the line, grab substring */ - result = pcre_exec(drweb_re, NULL, CS tmpbuf, Ustrlen(tmpbuf), 0, 0, - ovector, nelem(ovector)); + result = pcre2_match(drweb_re, (PCRE2_SPTR)tmpbuf, PCRE2_ZERO_TERMINATED, + 0, 0, md, pcre_gen_mtc_ctx); if (result >= 2) { - const char * pre_malware_nb; - - pcre_get_substring(CS tmpbuf, ovector, result, 1, &pre_malware_nb); + PCRE2_SIZE * ovec = pcre2_get_ovector_pointer(md); if (i==0) /* the first name we just copy to malware_name */ - g = string_cat(NULL, US pre_malware_nb); + g = string_catn(NULL, US ovec[2], ovec[3] - ovec[2]); - /*XXX could be string_append_listele? */ else /* concatenate each new virus name to previous */ - g = string_append(g, 2, "/", pre_malware_nb); - - pcre_free_substring(pre_malware_nb); + { + g = string_catn(g, US"/", 1); + g = string_catn(g, US ovec[2], ovec[3] - ovec[2]); + } } + /* pcre2_match_data_free(md); gen ctx needs no free */ } malware_name = string_from_gstring(g); } @@ -1048,7 +1056,7 @@ badseek: err = errno; #ifndef DISABLE_MAL_FSECURE case M_FSEC: /* "fsecure" scanner type ---------------------------------- */ { - int i, j, bread = 0; + int i, bread = 0; uschar * file_name; uschar av_buffer[1024]; static uschar *cmdopt[] = { US"CONFIGURE\tARCHIVE\t1\n", @@ -1068,13 +1076,13 @@ badseek: err = errno; if (m_sock_send(malware_daemon_ctx.sock, cmdopt[i], Ustrlen(cmdopt[i]), &errstr) < 0) return m_panic_defer(scanent, CUS callout_address, errstr); - bread = ip_recv(&malware_daemon_ctx, av_buffer, sizeof(av_buffer), tmo-time(NULL)); + bread = ip_recv(&malware_daemon_ctx, av_buffer, sizeof(av_buffer), tmo); if (bread > 0) av_buffer[bread]='\0'; if (bread < 0) return m_panic_defer_3(scanent, CUS callout_address, string_sprintf("unable to read answer %d (%s)", i, strerror(errno)), malware_daemon_ctx.sock); - for (j = 0; j < bread; j++) + for (int j = 0; j < bread; j++) if (av_buffer[j] == '\r' || av_buffer[j] == '\n') av_buffer[j] ='@'; } @@ -1088,7 +1096,7 @@ badseek: err = errno; /* set up match */ /* todo also SUSPICION\t */ if (!fsec_re) - fsec_re = m_pcre_compile(fsec_re_str, &errstr); + fsec_re = m_pcre_compile(fsec_re_str, FALSE, &errstr); /* read report, linewise. Apply a timeout as the Fsecure daemon sometimes wants an answer to "PING" but they won't tell us what */ @@ -1100,7 +1108,7 @@ badseek: err = errno; { errno = ETIMEDOUT; i = av_buffer+sizeof(av_buffer)-p; - if ((bread= ip_recv(&malware_daemon_ctx, p, i-1, tmo-time(NULL))) < 0) + if ((bread= ip_recv(&malware_daemon_ctx, p, i-1, tmo)) < 0) return m_panic_defer_3(scanent, CUS callout_address, string_sprintf("unable to read result (%s)", strerror(errno)), malware_daemon_ctx.sock); @@ -1139,7 +1147,7 @@ badseek: err = errno; int kav_rc; unsigned long kav_reportlen; int bread; - const pcre *kav_re; + const pcre2_code *kav_re; uschar *p; /* get current date and time, build scan request */ @@ -1212,12 +1220,12 @@ badseek: err = errno; /* set up match regex, depends on retcode */ if (kav_rc == 3) { - if (!kav_re_sus) kav_re_sus = m_pcre_compile(kav_re_sus_str, &errstr); + if (!kav_re_sus) kav_re_sus = m_pcre_compile(kav_re_sus_str, FALSE, &errstr); kav_re = kav_re_sus; } else { - if (!kav_re_inf) kav_re_inf = m_pcre_compile(kav_re_inf_str, &errstr); + if (!kav_re_inf) kav_re_inf = m_pcre_compile(kav_re_inf_str, FALSE, &errstr); kav_re = kav_re_inf; } @@ -1248,8 +1256,8 @@ badseek: err = errno; case M_CMDL: /* "cmdline" scanner type ---------------------------------- */ { const uschar *cmdline_scanner = scanner_options; - const pcre *cmdline_trigger_re; - const pcre *cmdline_regex_re; + const pcre2_code *cmdline_trigger_re; + const pcre2_code *cmdline_regex_re; uschar * file_name; uschar * commandline; void (*eximsigchld)(int); @@ -1266,13 +1274,13 @@ badseek: err = errno; return m_panic_defer(scanent, NULL, errstr); /* find scanner output trigger */ - cmdline_trigger_re = m_pcre_nextinlist(&av_scanner_work, &sep, + cmdline_trigger_re = m_pcre_nextinlist(&av_scanner_work, &sep, av_scanner_textonly, "missing trigger specification", &errstr); if (!cmdline_trigger_re) return m_panic_defer(scanent, NULL, errstr); /* find scanner name regex */ - cmdline_regex_re = m_pcre_nextinlist(&av_scanner_work, &sep, + cmdline_regex_re = m_pcre_nextinlist(&av_scanner_work, &sep, av_scanner_textonly, "missing virus name regex specification", &errstr); if (!cmdline_regex_re) return m_panic_defer(scanent, NULL, errstr); @@ -1365,13 +1373,10 @@ badseek: err = errno; malware_name = US"unknown"; /* re-open the scanner output file, look for name match */ - scanner_record = fopen(CS file_name, "rb"); - while (fgets(CS linebuffer, sizeof(linebuffer), scanner_record)) - { - /* try match */ - if ((s = m_pcre_exec(cmdline_regex_re, linebuffer))) + scanner_record = Ufopen(file_name, "rb"); + while (Ufgets(linebuffer, sizeof(linebuffer), scanner_record)) + if ((s = m_pcre_exec(cmdline_regex_re, linebuffer))) /* try match */ malware_name = s; - } (void)fclose(scanner_record); } else /* no virus found */ @@ -1405,7 +1410,7 @@ badseek: err = errno; /* wait for result */ memset(av_buffer, 0, sizeof(av_buffer)); - if ((bread = ip_recv(&malware_daemon_ctx, av_buffer, sizeof(av_buffer), tmo-time(NULL))) <= 0) + if ((bread = ip_recv(&malware_daemon_ctx, av_buffer, sizeof(av_buffer), tmo)) <= 0) return m_panic_defer_3(scanent, CUS callout_address, string_sprintf("unable to read from UNIX socket (%s)", scanner_options), malware_daemon_ctx.sock); @@ -1446,9 +1451,7 @@ badseek: err = errno; uschar av_buffer[1024]; uschar *hostname = US""; host_item connhost; - uschar *clamav_fbuf; - int clam_fd, result; - off_t fsize; + int clam_fd; unsigned int fsize_uint; BOOL use_scan_command = FALSE; clamd_address * cv[MAX_CLAMD_SERVERS]; @@ -1466,9 +1469,9 @@ badseek: err = errno; int subsep = ' '; /* Local file; so we def want to use_scan_command and don't want to try - * passing IP/port combinations */ + passing IP/port combinations */ use_scan_command = TRUE; - cd = (clamd_address *) store_get(sizeof(clamd_address)); + cd = (clamd_address *) store_get(sizeof(clamd_address), GET_UNTAINTED); /* extract socket-path part */ sublist = scanner_options; @@ -1502,7 +1505,7 @@ badseek: err = errno; continue; } - cd = (clamd_address *) store_get(sizeof(clamd_address)); + cd = (clamd_address *) store_get(sizeof(clamd_address), GET_UNTAINTED); /* extract host and port part */ sublist = scanner_options; @@ -1555,16 +1558,17 @@ badseek: err = errno; { cmd_str.data = US"zINSTREAM"; cmd_str.len = 10; } else { - cmd_str.data = string_sprintf("SCAN %s\n", eml_filename); - cmd_str.len = Ustrlen(cmd_str.data); + int n; + cmd_str.data = string_sprintf("SCAN %s\n%n", eml_filename, &n); + cmd_str.len = n; /* .len is a size_t */ } /* We have some network servers specified */ if (num_servers) { /* Confirmed in ClamAV source (0.95.3) that the TCPAddr option of clamd - * only supports AF_INET, but we should probably be looking to the - * future and rewriting this to be protocol-independent anyway. */ + only supports AF_INET, but we should probably be looking to the + future and rewriting this to be protocol-independent anyway. */ while (num_servers > 0) { @@ -1575,15 +1579,17 @@ badseek: err = errno; cd->hostspec, cd->tcp_port); /* Lookup the host. This is to ensure that we connect to the same IP - * on both connections (as one host could resolve to multiple ips) */ + on both connections (as one host could resolve to multiple ips) */ for (;;) { + /*XXX we trust that the cmd_str is idempotent */ if ((malware_daemon_ctx.sock = m_tcpsocket(cd->hostspec, cd->tcp_port, - &connhost, &errstr, &cmd_str)) >= 0) + &connhost, &errstr, + use_scan_command ? &cmd_str : NULL)) >= 0) { /* Connection successfully established with a server */ hostname = cd->hostspec; - cmd_str.len = 0; + if (use_scan_command) cmd_str.len = 0; break; } if (cd->retry <= 0) break; @@ -1617,20 +1623,28 @@ badseek: err = errno; } /* have socket in variable "sock"; command to use is semi-independent of - * the socket protocol. We use SCAN if is local (either Unix/local - * domain socket, or explicitly told local) else we stream the data. - * How we stream the data depends upon how we were built. */ + the socket protocol. We use SCAN if is local (either Unix/local + domain socket, or explicitly told local) else we stream the data. + How we stream the data depends upon how we were built. */ if (!use_scan_command) { + struct stat st; +#if defined(EXIM_TCP_CORK) && !defined(OS_SENDFILE) + BOOL corked = TRUE; +#endif /* New protocol: "zINSTREAM\n" followed by a sequence of chunks, a 4-byte number (network order), terminated by a zero-length - chunk. */ + chunk. We only send one chunk. */ DEBUG(D_acl) debug_printf_indent( "Malware scan: issuing %s new-style remote scan (zINSTREAM)\n", scanner_name); +#if defined(EXIM_TCP_CORK) + (void) setsockopt(malware_daemon_ctx.sock, IPPROTO_TCP, EXIM_TCP_CORK, + US &on, sizeof(on)); +#endif /* Pass the string to ClamAV (10 = "zINSTREAM\0"), if not already sent */ if (cmd_str.len) if (send(malware_daemon_ctx.sock, cmd_str.data, cmd_str.len, 0) < 0) @@ -1639,8 +1653,7 @@ badseek: err = errno; strerror(errno)), malware_daemon_ctx.sock); - /* calc file size */ - if ((clam_fd = open(CS eml_filename, O_RDONLY)) < 0) + if ((clam_fd = exim_open2(CS eml_filename, O_RDONLY)) < 0) { int err = errno; return m_panic_defer_3(scanent, NULL, @@ -1648,67 +1661,79 @@ badseek: err = errno; eml_filename, strerror(err)), malware_daemon_ctx.sock); } - if ((fsize = lseek(clam_fd, 0, SEEK_END)) < 0) + if (fstat(clam_fd, &st) < 0) { - int err; -b_seek: err = errno; + int err = errno; (void)close(clam_fd); return m_panic_defer_3(scanent, NULL, - string_sprintf("can't seek spool file %s: %s", + string_sprintf("can't stat spool file %s: %s", eml_filename, strerror(err)), malware_daemon_ctx.sock); } - fsize_uint = (unsigned int) fsize; - if ((off_t)fsize_uint != fsize) + fsize_uint = (unsigned int) st.st_size; + if ((off_t)fsize_uint != st.st_size) { (void)close(clam_fd); return m_panic_defer_3(scanent, NULL, - string_sprintf("seeking spool file %s, size overflow", - eml_filename), + string_sprintf("stat spool file %s, size overflow", eml_filename), malware_daemon_ctx.sock); } - if (lseek(clam_fd, 0, SEEK_SET) < 0) - goto b_seek; - if (!(clamav_fbuf = US malloc(fsize_uint))) - { - (void)close(clam_fd); + /* send file size */ + send_size = htonl(fsize_uint); + if (send(malware_daemon_ctx.sock, &send_size, sizeof(send_size), 0) < 0) return m_panic_defer_3(scanent, NULL, - string_sprintf("unable to allocate memory %u for file (%s)", - fsize_uint, eml_filename), + string_sprintf("unable to send file size to socket (%s)", hostname), malware_daemon_ctx.sock); - } - if ((result = read(clam_fd, clamav_fbuf, fsize_uint)) < 0) + /* send file body */ + while (fsize_uint) { - int err = errno; - free(clamav_fbuf); (void)close(clam_fd); - return m_panic_defer_3(scanent, NULL, - string_sprintf("can't read spool file %s: %s", - eml_filename, strerror(err)), - malware_daemon_ctx.sock); +#ifdef OS_SENDFILE + int n = os_sendfile(malware_daemon_ctx.sock, clam_fd, NULL, (size_t)fsize_uint); + if (n < 0) + return m_panic_defer_3(scanent, NULL, + string_sprintf("unable to send file body to socket (%s): %s", hostname, strerror(errno)), + malware_daemon_ctx.sock); + fsize_uint -= n; +#else + int n = MIN(fsize_uint, big_buffer_size); + if ((n = read(clam_fd, big_buffer, n)) < 0) + return m_panic_defer_3(scanent, NULL, + string_sprintf("can't read spool file %s: %s", + eml_filename, strerror(errno)), + malware_daemon_ctx.sock); + if (send(malware_daemon_ctx.sock, big_buffer, (size_t)n, 0) < 0) + return m_panic_defer_3(scanent, NULL, + string_sprintf("unable to send file body to socket (%s): %s", hostname, strerror(errno)), + malware_daemon_ctx.sock); + fsize_uint -= n; +# ifdef EXIM_TCP_CORK + if (corked) + { + corked = FALSE; + (void) setsockopt(malware_daemon_ctx.sock, IPPROTO_TCP, EXIM_TCP_CORK, + US &off, sizeof(off)); + } +# endif +#endif /*!OS_SENDFILE*/ + } - (void)close(clam_fd); - /* send file body to socket */ - send_size = htonl(fsize_uint); send_final_zeroblock = 0; - if ((send(malware_daemon_ctx.sock, &send_size, sizeof(send_size), 0) < 0) || - (send(malware_daemon_ctx.sock, clamav_fbuf, fsize_uint, 0) < 0) || - (send(malware_daemon_ctx.sock, &send_final_zeroblock, sizeof(send_final_zeroblock), 0) < 0)) - { - free(clamav_fbuf); + if (send(malware_daemon_ctx.sock, &send_final_zeroblock, sizeof(send_final_zeroblock), 0) < 0) return m_panic_defer_3(scanent, NULL, - string_sprintf("unable to send file body to socket (%s)", hostname), + string_sprintf("unable to send file terminator to socket (%s)", hostname), malware_daemon_ctx.sock); - } - - free(clamav_fbuf); +#ifdef OS_SENDFILE + (void) setsockopt(malware_daemon_ctx.sock, IPPROTO_TCP, EXIM_TCP_CORK, + US &off, sizeof(off)); +#endif } else { /* use scan command */ /* Send a SCAN command pointing to a filename; then in the then in the - * scan-method-neutral part, read the response back */ + scan-method-neutral part, read the response back */ /* ================================================================= */ @@ -1733,14 +1758,14 @@ b_seek: err = errno; malware_daemon_ctx.sock); /* Do not shut down the socket for writing; a user report noted that - * clamd 0.70 does not react well to this. */ + clamd 0.70 does not react well to this. */ } /* Commands have been sent, no matter which scan method or connection - * type we're using; now just read the result, independent of method. */ + type we're using; now just read the result, independent of method. */ /* Read the result */ memset(av_buffer, 0, sizeof(av_buffer)); - bread = ip_recv(&malware_daemon_ctx, av_buffer, sizeof(av_buffer), tmo-time(NULL)); + bread = ip_recv(&malware_daemon_ctx, av_buffer, sizeof(av_buffer), tmo); (void)close(malware_daemon_ctx.sock); malware_daemon_ctx.sock = -1; malware_daemon_ctx.tls_ctx = NULL; @@ -1782,6 +1807,7 @@ b_seek: err = errno; /* strip newline at the end (won't be present for zINSTREAM) (also any trailing whitespace, which shouldn't exist, but we depend upon this below, so double-check) */ + p = av_buffer + Ustrlen(av_buffer) - 1; if (*p == '\n') *p = '\0'; @@ -1792,7 +1818,7 @@ b_seek: err = errno; if (*p) ++p; /* colon in returned output? */ - if(!(p = Ustrchr(av_buffer,':'))) + if (!(p = Ustrchr(av_buffer,':'))) return m_panic_defer(scanent, CUS callout_address, string_sprintf( "ClamAV returned malformed result (missing colon): %s", av_buffer)); @@ -1856,8 +1882,8 @@ b_seek: err = errno; uschar * linebuffer; uschar * sockline_scanner; uschar sockline_scanner_default[] = "%s\n"; - const pcre *sockline_trig_re; - const pcre *sockline_name_re; + const pcre2_code *sockline_trig_re; + const pcre2_code *sockline_name_re; /* find scanner command line */ if ( (sockline_scanner = string_nextinlist(&av_scanner_work, &sep, @@ -1877,13 +1903,13 @@ b_seek: err = errno; string_printing(sockline_scanner)); /* find scanner output trigger */ - sockline_trig_re = m_pcre_nextinlist(&av_scanner_work, &sep, + sockline_trig_re = m_pcre_nextinlist(&av_scanner_work, &sep, av_scanner_textonly, "missing trigger specification", &errstr); if (!sockline_trig_re) return m_panic_defer_3(scanent, NULL, errstr, malware_daemon_ctx.sock); /* find virus name regex */ - sockline_name_re = m_pcre_nextinlist(&av_scanner_work, &sep, + sockline_name_re = m_pcre_nextinlist(&av_scanner_work, &sep, av_scanner_textonly, "missing virus name regex specification", &errstr); if (!sockline_name_re) return m_panic_defer_3(scanent, NULL, errstr, malware_daemon_ctx.sock); @@ -1898,7 +1924,7 @@ b_seek: err = errno; return m_panic_defer(scanent, CUS callout_address, errstr); /* Read the result */ - bread = ip_recv(&malware_daemon_ctx, av_buffer, sizeof(av_buffer), tmo-time(NULL)); + bread = ip_recv(&malware_daemon_ctx, av_buffer, sizeof(av_buffer), tmo); if (bread <= 0) return m_panic_defer_3(scanent, CUS callout_address, @@ -1980,7 +2006,7 @@ b_seek: err = errno; and the [ ] marker. [+] - not infected [L] - infected - [E] - some error occured + [E] - some error occurred Such marker follows the first non-escaped TAB. For more information see avast-protocol(5) @@ -2014,11 +2040,11 @@ b_seek: err = errno; */ if ( ( !ava_re_clean - && !(ava_re_clean = m_pcre_compile(ava_re_clean_str, &errstr))) + && !(ava_re_clean = m_pcre_compile(ava_re_clean_str, FALSE, &errstr))) || ( !ava_re_virus - && !(ava_re_virus = m_pcre_compile(ava_re_virus_str, &errstr))) + && !(ava_re_virus = m_pcre_compile(ava_re_virus_str, FALSE, &errstr))) || ( !ava_re_error - && !(ava_re_error = m_pcre_compile(ava_re_error_str, &errstr))) + && !(ava_re_error = m_pcre_compile(ava_re_error_str, FALSE, &errstr))) ) return malware_panic_defer(errstr); @@ -2096,7 +2122,7 @@ b_seek: err = errno; if (malware_name) /* Nothing else matters, just read on */ break; - if (pcre_exec(ava_re_clean, NULL, CS buf, slen, 0, 0, NULL, 0) == 0) + if (regex_match(ava_re_clean, buf, slen, NULL)) break; if ((malware_name = m_pcre_exec(ava_re_virus, buf))) @@ -2117,7 +2143,7 @@ b_seek: err = errno; break; } } - else if (pcre_exec(ava_re_error, NULL, CS buf, slen, 0, 0, NULL, 0) == 0) + else if (regex_match(ava_re_error, buf, slen, NULL)) { log_write(0, LOG_MAIN, "internal scanner error (ignored): %s", buf); break; @@ -2180,15 +2206,16 @@ filename; it's a wrapper around the malware_file function. Arguments: malware_re match condition for "malware=" + cacheable the RE did not use any dynamic elements during expansion timeout if nonzero, timeout in seconds Returns: Exim message processing code (OK, FAIL, DEFER, ...) where true means malware was found (condition applies) */ int -malware(const uschar * malware_re, int timeout) +malware(const uschar * malware_re, BOOL cacheable, int timeout) { -int ret = malware_internal(malware_re, NULL, timeout); +int ret = malware_internal(malware_re, cacheable, NULL, timeout); if (ret == DEFER) av_failed = TRUE; return ret; @@ -2226,9 +2253,9 @@ sender_address = US"malware-sender@example.net"; return_path = US""; recipients_list = NULL; receive_add_recipient(US"malware-victim@example.net", -1); -enable_dollar_recipients = TRUE; +f.enable_dollar_recipients = TRUE; -ret = malware_internal(US"*", eml_filename, 0); +ret = malware_internal(US"*", TRUE, eml_filename, 0); Ustrncpy(spooled_message_id, message_id, sizeof(spooled_message_id)); spool_mbox_ok = 1; @@ -2249,46 +2276,46 @@ void malware_init(void) { if (!malware_default_re) - malware_default_re = regex_must_compile(malware_regex_default, FALSE, TRUE); + malware_default_re = regex_must_compile(malware_regex_default, MCS_NOFLAGS, TRUE); #ifndef DISABLE_MAL_DRWEB if (!drweb_re) - drweb_re = regex_must_compile(drweb_re_str, FALSE, TRUE); + drweb_re = regex_must_compile(drweb_re_str, MCS_NOFLAGS, TRUE); #endif #ifndef DISABLE_MAL_FSECURE if (!fsec_re) - fsec_re = regex_must_compile(fsec_re_str, FALSE, TRUE); + fsec_re = regex_must_compile(fsec_re_str, MCS_NOFLAGS, TRUE); #endif #ifndef DISABLE_MAL_KAV if (!kav_re_sus) - kav_re_sus = regex_must_compile(kav_re_sus_str, FALSE, TRUE); + kav_re_sus = regex_must_compile(kav_re_sus_str, MCS_NOFLAGS, TRUE); if (!kav_re_inf) - kav_re_inf = regex_must_compile(kav_re_inf_str, FALSE, TRUE); + kav_re_inf = regex_must_compile(kav_re_inf_str, MCS_NOFLAGS, TRUE); #endif #ifndef DISABLE_MAL_AVAST if (!ava_re_clean) - ava_re_clean = regex_must_compile(ava_re_clean_str, FALSE, TRUE); + ava_re_clean = regex_must_compile(ava_re_clean_str, MCS_NOFLAGS, TRUE); if (!ava_re_virus) - ava_re_virus = regex_must_compile(ava_re_virus_str, FALSE, TRUE); + ava_re_virus = regex_must_compile(ava_re_virus_str, MCS_NOFLAGS, TRUE); if (!ava_re_error) - ava_re_error = regex_must_compile(ava_re_error_str, FALSE, TRUE); + ava_re_error = regex_must_compile(ava_re_error_str, MCS_NOFLAGS, TRUE); #endif #ifndef DISABLE_MAL_FFROT6D if (!fprot6d_re_error) - fprot6d_re_error = regex_must_compile(fprot6d_re_error_str, FALSE, TRUE); + fprot6d_re_error = regex_must_compile(fprot6d_re_error_str, MCS_NOFLAGS, TRUE); if (!fprot6d_re_virus) - fprot6d_re_virus = regex_must_compile(fprot6d_re_virus_str, FALSE, TRUE); + fprot6d_re_virus = regex_must_compile(fprot6d_re_virus_str, MCS_NOFLAGS, TRUE); #endif } -void -malware_show_supported(FILE * f) +gstring * +malware_show_supported(gstring * g) { -struct scan * sc; -fprintf(f, "Malware:"); -for (sc = m_scans; sc->scancode != -1; sc++) fprintf(f, " %s", sc->name); -fprintf(f, "\n"); +g = string_cat(g, US"Malware:"); +for (struct scan * sc = m_scans; sc->scancode != (scanner_t)-1; sc++) + g = string_fmt_append(g, " %s", sc->name); +return string_cat(g, US"\n"); }