1
0
mirror of https://github.com/nmap/nmap.git synced 2025-12-06 04:31:29 +00:00
Files
nmap/tcpip.cc
2006-01-22 22:05:26 +00:00

2949 lines
97 KiB
C++

/***************************************************************************
* tcpip.cc -- Various functions relating to low level TCP/IP handling, *
* including sending raw packets, routing, printing packets, reading from *
* libpcap, etc. *
* *
***********************IMPORTANT NMAP LICENSE TERMS************************
* *
* The Nmap Security Scanner is (C) 1996-2004 Insecure.Com LLC. Nmap *
* is also a registered trademark of Insecure.Com LLC. This program is *
* free software; you may redistribute and/or modify it under the *
* terms of the GNU General Public License as published by the Free *
* Software Foundation; Version 2. This guarantees your right to use, *
* modify, and redistribute this software under certain conditions. If *
* you wish to embed Nmap technology into proprietary software, we may be *
* willing to sell alternative licenses (contact sales@insecure.com). *
* Many security scanner vendors already license Nmap technology such as *
* our remote OS fingerprinting database and code, service/version *
* detection system, and port scanning code. *
* *
* Note that the GPL places important restrictions on "derived works", yet *
* it does not provide a detailed definition of that term. To avoid *
* misunderstandings, we consider an application to constitute a *
* "derivative work" for the purpose of this license if it does any of the *
* following: *
* o Integrates source code from Nmap *
* o Reads or includes Nmap copyrighted data files, such as *
* nmap-os-fingerprints or nmap-service-probes. *
* o Executes Nmap and parses the results (as opposed to typical shell or *
* execution-menu apps, which simply display raw Nmap output and so are *
* not derivative works.) *
* o Integrates/includes/aggregates Nmap into a proprietary executable *
* installer, such as those produced by InstallShield. *
* o Links to a library or executes a program that does any of the above *
* *
* The term "Nmap" should be taken to also include any portions or derived *
* works of Nmap. This list is not exclusive, but is just meant to *
* clarify our interpretation of derived works with some common examples. *
* These restrictions only apply when you actually redistribute Nmap. For *
* example, nothing stops you from writing and selling a proprietary *
* front-end to Nmap. Just distribute it by itself, and point people to *
* http://www.insecure.org/nmap/ to download Nmap. *
* *
* We don't consider these to be added restrictions on top of the GPL, but *
* just a clarification of how we interpret "derived works" as it applies *
* to our GPL-licensed Nmap product. This is similar to the way Linus *
* Torvalds has announced his interpretation of how "derived works" *
* applies to Linux kernel modules. Our interpretation refers only to *
* Nmap - we don't speak for any other GPL products. *
* *
* If you have any questions about the GPL licensing restrictions on using *
* Nmap in non-GPL works, we would be happy to help. As mentioned above, *
* we also offer alternative license to integrate Nmap into proprietary *
* applications and appliances. These contracts have been sold to many *
* security vendors, and generally include a perpetual license as well as *
* providing for priority support and updates as well as helping to fund *
* the continued development of Nmap technology. Please email *
* sales@insecure.com for further information. *
* *
* As a special exception to the GPL terms, Insecure.Com LLC grants *
* permission to link the code of this program with any version of the *
* OpenSSL library which is distributed under a license identical to that *
* listed in the included Copying.OpenSSL file, and distribute linked *
* combinations including the two. You must obey the GNU GPL in all *
* respects for all of the code used other than OpenSSL. If you modify *
* this file, you may extend this exception to your version of the file, *
* but you are not obligated to do so. *
* *
* If you received these files with a written license agreement or *
* contract stating terms other than the terms above, then that *
* alternative license agreement takes precedence over these comments. *
* *
* Source is provided to this software because we believe users have a *
* right to know exactly what a program is going to do before they run it. *
* This also allows you to audit the software for security holes (none *
* have been found so far). *
* *
* Source code also allows you to port Nmap to new platforms, fix bugs, *
* and add new features. You are highly encouraged to send your changes *
* to fyodor@insecure.org for possible incorporation into the main *
* distribution. By sending these changes to Fyodor or one the *
* Insecure.Org development mailing lists, it is assumed that you are *
* offering Fyodor and Insecure.Com LLC the unlimited, non-exclusive right *
* to reuse, modify, and relicense the code. Nmap will always be *
* available Open Source, but this is important because the inability to *
* relicense code has caused devastating problems for other Free Software *
* projects (such as KDE and NASM). We also occasionally relicense the *
* code to third parties as discussed above. If you wish to specify *
* special license conditions of your contributions, just say so when you *
* send them. *
* *
* This program is distributed in the hope that it will be useful, but *
* WITHOUT ANY WARRANTY; without even the implied warranty of *
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU *
* General Public License for more details at *
* http://www.gnu.org/copyleft/gpl.html , or in the COPYING file included *
* with Nmap. *
* *
***************************************************************************/
/* $Id$ */
#include <dnet.h>
#include "tcpip.h"
#include "NmapOps.h"
#if HAVE_SYS_TIME_H
#include <sys/time.h>
#endif
#if HAVE_SYS_RESOURCE_H
#include <sys/resource.h>
#endif
#if HAVE_UNISTD_H
/* #include <sys/unistd.h> */
#include <unistd.h>
#endif
#if HAVE_NET_IF_H
#ifndef NET_IF_H /* why doesn't OpenBSD do this? */
#include <net/if.h>
#define NET_IF_H
#endif
#endif
#if HAVE_NETINET_IF_ETHER_H
#ifndef NETINET_IF_ETHER_H
#include <netinet/if_ether.h>
#define NETINET_IF_ETHER_H
#endif /* NETINET_IF_ETHER_H */
#endif /* HAVE_NETINET_IF_ETHER_H */
extern NmapOps o;
#ifdef __amigaos__
extern void CloseLibs(void);
#endif
#ifdef WIN32
#include "pcap-int.h"
void nmapwin_init();
void nmapwin_cleanup();
void nmapwin_list_interfaces();
int if2nameindex(int ifi);
#endif
static PacketCounter PktCt;
void sethdrinclude(int sd) {
#ifdef IP_HDRINCL
int one = 1;
setsockopt(sd, IPPROTO_IP, IP_HDRINCL, (const char *) &one, sizeof(one));
#endif
}
// Takes a protocol number like IPPROTO_TCP, IPPROTO_UDP, or
// IPPROTO_TCP and returns a ascii representation (or "unknown" if it
// doesn't recognize the number). If uppercase is true, the returned
// value will be in all uppercase letters. You can skip this
// parameter to use lowercase.
const char *proto2ascii(u8 proto, bool uppercase) {
switch(proto) {
case IPPROTO_TCP:
return uppercase? "TCP" : "tcp"; break;
case IPPROTO_UDP:
return uppercase? "UDP" : "udp"; break;
case IPPROTO_IP:
return uppercase? "IP" : "ip"; break;
default:
return uppercase? "UNKNOWN" : "unknown";
}
return NULL; // Unreached
}
static char *ll2shortascii(unsigned long long bytes, char *buf, int buflen) {
if (buflen < 2 || !buf) fatal("Bogus parameter passed to ll2shortascii");
if (bytes > 1000000) {
snprintf(buf, buflen, "%.3gMB", bytes / 1000000.0);
} else if (bytes > 10000) {
snprintf(buf, buflen, "%.3gKB", bytes / 1000.0);
} else snprintf(buf, buflen, "%uB", (unsigned int) bytes);
return buf;
}
/* Fill buf (up to buflen -- truncate if necessary but always
terminate) with a short representation of the packet stats.
Returns buf. Aborts if there is a problem. */
char *getFinalPacketStats(char *buf, int buflen) {
char sendbytesasc[16], recvbytesasc[16];
if (buflen <= 10 || !buf)
fatal("getFinalPacketStats called with woefully inadequate parameters");
snprintf(buf, buflen,
#if WIN32
"Raw packets sent: %I64u (%s) | Rcvd: %I64u (%s)",
#else
"Raw packets sent: %llu (%s) | Rcvd: %llu (%s)",
#endif
PktCt.sendPackets,
ll2shortascii(PktCt.sendBytes, sendbytesasc, sizeof(sendbytesasc)),
PktCt.recvPackets,
ll2shortascii(PktCt.recvBytes, recvbytesasc, sizeof(recvbytesasc)));
return buf;
}
/* Takes an ARP PACKET (including ethernet header) and prints it if
packet tracing is enabled. 'frame' must point to the 14-byte
ethernet header (e.g. starting with destination addr). The
direction must be PacketTrace::SENT or PacketTrace::RCVD .
Optional 'now' argument makes this function slightly more
efficient by avoiding a gettimeofday() call. */
void PacketTrace::traceArp(pdirection pdir, const u8 *frame, u32 len,
struct timeval *now) {
struct timeval tv;
char arpdesc[128];
char who_has[INET_ADDRSTRLEN], tell[INET_ADDRSTRLEN];
if (pdir == SENT) {
PktCt.sendPackets++;
PktCt.sendBytes += len;
} else {
PktCt.recvPackets++;
PktCt.recvBytes += len;
}
if (!o.packetTrace()) return;
if (now)
tv = *now;
else gettimeofday(&tv, NULL);
if (len < 42) {
error("Packet tracer: Arp packets must be at least 42 bytes long. Should be exactly that length excl. ethernet padding.");
return;
}
if (frame[21] == 1) /* arp REQUEST */ {
inet_ntop(AF_INET, frame+38, who_has, sizeof(who_has));
inet_ntop(AF_INET, frame+28, tell, sizeof(who_has));
snprintf(arpdesc, sizeof(arpdesc), "who-has %s tell %s", who_has, tell);
} else { /* ARP REPLY */
inet_ntop(AF_INET, frame+28, who_has, sizeof(who_has));
snprintf(arpdesc, sizeof(arpdesc),
"reply %s is-at %02X:%02X:%02X:%02X:%02X:%02X", who_has,
frame[22], frame[23], frame[24], frame[25], frame[26], frame[27]);
}
log_write(LOG_STDOUT|LOG_NORMAL, "%s (%.4fs) ARP %s\n", (pdir == SENT)? "SENT" : "RCVD", o.TimeSinceStartMS(&tv) / 1000.0, arpdesc);
return;
}
/* Takes an IP PACKET and prints it if packet tracing is enabled.
'packet' must point to the IPv4 header. The direction must be
PacketTrace::SENT or PacketTrace::RCVD . Optional 'now' argument
makes this function slightly more efficient by avoiding a gettimeofday()
call. */
void PacketTrace::trace(pdirection pdir, const u8 *packet, u32 len,
struct timeval *now) {
struct timeval tv;
if (pdir == SENT) {
PktCt.sendPackets++;
PktCt.sendBytes += len;
} else {
PktCt.recvPackets++;
PktCt.recvBytes += len;
}
if (!o.packetTrace()) return;
if (now)
tv = *now;
else gettimeofday(&tv, NULL);
if (len < 20) {
error("Packet tracer: tiny packet encountered");
return;
}
log_write(LOG_STDOUT|LOG_NORMAL, "%s (%.4fs) %s\n", (pdir == SENT)? "SENT" : "RCVD", o.TimeSinceStartMS(&tv) / 1000.0, ippackethdrinfo(packet, len));
return;
}
/* Adds a trace entry when a connect() is attempted if packet tracing
is enabled. Pass IPPROTO_TCP or IPPROTO_UDP as the protocol. The
sock may be a sockaddr_in or sockaddr_in6. The return code of
connect is passed in connectrc. If the return code is -1, get the
errno and pass that as connect_errno. */
void PacketTrace::traceConnect(u8 proto, const struct sockaddr *sock,
int socklen, int connectrc, int connect_errno,
const struct timeval *now) {
struct sockaddr_in *sin = (struct sockaddr_in *) sock;
#if HAVE_IPV6
struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) sock;
#endif
struct timeval tv;
char errbuf[64] = "";
char targetipstr[INET6_ADDRSTRLEN] = "";
u16 targetport = 0;
if (!o.packetTrace()) return;
if (now)
tv = *now;
else gettimeofday(&tv, NULL);
assert(proto == IPPROTO_TCP || proto == IPPROTO_UDP);
if (connectrc == 0)
Strncpy(errbuf, "Connected", sizeof(errbuf));
else {
snprintf(errbuf, sizeof(errbuf), "%s", strerror(connect_errno));
}
if (sin->sin_family == AF_INET) {
if (inet_ntop(sin->sin_family, (char *) &sin->sin_addr, targetipstr,
sizeof(targetipstr)) == NULL)
fatal("Failed to convert target IPv4 address to presentation format!?!");
targetport = ntohs(sin->sin_port);
} else {
#if HAVE_IPV6
assert(sin->sin_family == AF_INET6);
if (inet_ntop(sin->sin_family, (char *) &sin6->sin6_addr, targetipstr,
sizeof(targetipstr)) == NULL)
fatal("Failed to convert target IPv4 address to presentation format!?!");
targetport = ntohs(sin6->sin6_port);
#else
assert(0);
#endif
}
log_write(LOG_STDOUT|LOG_NORMAL, "CONN (%.4fs) %s localhost > %s:%d => %s\n",
o.TimeSinceStartMS(&tv) / 1000.0,
(proto == IPPROTO_TCP)? "TCP" : "UDP", targetipstr, targetport,
errbuf);
}
/* Converts an IP address given in a sockaddr_storage to an IPv4 or
IPv6 IP address string. Since a static buffer is returned, this is
not thread-safe and can only be used once in calls like printf()
*/
const char *inet_socktop(struct sockaddr_storage *ss) {
static char buf[INET6_ADDRSTRLEN];
struct sockaddr_in *sin = (struct sockaddr_in *) ss;
#if HAVE_IPV6
struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) ss;
#endif
if (inet_ntop(sin->sin_family, (sin->sin_family == AF_INET)?
(char *) &sin->sin_addr :
#if HAVE_IPV6
(char *) &sin6->sin6_addr,
#else
(char *) NULL,
#endif /* HAVE_IPV6 */
buf, sizeof(buf)) == NULL) {
fatal("Failed to convert target address to presentation format in inet_socktop!?! Error: %s", strerror(socket_errno()));
}
return buf;
}
/* Tries to resolve the given name (or literal IP) into a sockaddr
structure. The af should be PF_INET (for IPv4) or PF_INET6. Returns 0
if hostname cannot be resolved. It is OK to pass in a sockaddr_in or
sockaddr_in6 casted to a sockaddr_storage as long as you use the matching
pf.*/
int resolve(char *hostname, struct sockaddr_storage *ss, size_t *sslen,
int pf) {
struct addrinfo hints;
struct addrinfo *result;
int rc;
assert(ss);
assert(sslen);
memset(&hints, 0, sizeof(hints));
hints.ai_family = pf;
rc = getaddrinfo(hostname, NULL, &hints, &result);
if (rc != 0)
return 0;
assert(result->ai_addrlen > 0 && result->ai_addrlen <= (int) sizeof(struct sockaddr_storage));
*sslen = result->ai_addrlen;
memcpy(ss, result->ai_addr, *sslen);
freeaddrinfo(result);
return 1;
}
/* Returns a buffer of ASCII information about a packet that may look
like "TCP 127.0.0.1:50923 > 127.0.0.1:3 S ttl=61 id=39516 iplen=40
seq=625950769" or "ICMP PING (0/1) ttl=61 id=39516 iplen=40".
Since this is a static buffer, don't use threads or call twice
within (say) printf(). And certainly don't try to free() it! The
returned buffer is NUL-terminated */
const char *ippackethdrinfo(const u8 *packet, u32 len) {
static char protoinfo[256];
struct ip *ip = (struct ip *) packet;
struct tcphdr *tcp;
udphdr_bsd *udp;
char ipinfo[64];
char srchost[INET6_ADDRSTRLEN], dsthost[INET6_ADDRSTRLEN];
char *p;
struct in_addr saddr, daddr;
int frag_off = 0, more_fragments = 0;
char fragnfo[64] = "";
char tflags[10];
if (ip->ip_v != 4)
return "BOGUS! IP Version in packet is not 4";
if (len < sizeof(struct ip))
return "BOGUS! Packet too short.";
saddr.s_addr = ip->ip_src.s_addr;
daddr.s_addr = ip->ip_dst.s_addr;
inet_ntop(AF_INET, &saddr, srchost, sizeof(srchost));
inet_ntop(AF_INET, &daddr, dsthost, sizeof(dsthost));
frag_off = 8 * (ntohs(ip->ip_off) & 8191) /* 2^13 - 1 */;
more_fragments = ntohs(ip->ip_off) & IP_MF;
if (frag_off || more_fragments) {
snprintf(fragnfo, sizeof(fragnfo), " frag offset=%d%s", frag_off, more_fragments ? "+" : "");
}
snprintf(ipinfo, sizeof(ipinfo), "ttl=%d id=%d iplen=%d%s",
ip->ip_ttl, ntohs(ip->ip_id), ntohs(ip->ip_len), fragnfo);
if (ip->ip_p == IPPROTO_TCP) {
char tcpinfo[64] = "";
char buf[32];
tcp = (struct tcphdr *) (packet + ip->ip_hl * 4);
if (frag_off > 8 || len < (u32) ip->ip_hl * 4 + 8)
snprintf(protoinfo, sizeof(protoinfo), "TCP %s:?? > %s:?? ?? %s (incomplete)", srchost, dsthost, ipinfo);
else if (frag_off == 8 && len >= (u32) ip->ip_hl * 4 + 8) {// we can get TCP flags and ACKn
tcp = (struct tcphdr *)((u8 *) tcp - frag_off); // ugly?
p = tflags;
/* These are basically in tcpdump order */
if (tcp->th_flags & TH_SYN) *p++ = 'S';
if (tcp->th_flags & TH_FIN) *p++ = 'F';
if (tcp->th_flags & TH_RST) *p++ = 'R';
if (tcp->th_flags & TH_PUSH) *p++ = 'P';
if (tcp->th_flags & TH_ACK) {
*p++ = 'A';
snprintf(tcpinfo, sizeof(tcpinfo), " ack=%lu",
(unsigned long) ntohl(tcp->th_ack));
}
if (tcp->th_flags & TH_URG) *p++ = 'U';
if (tcp->th_flags & TH_ECE) *p++ = 'E'; /* rfc 2481/3168 */
if (tcp->th_flags & TH_CWR) *p++ = 'C'; /* rfc 2481/3168 */
*p++ = '\0';
snprintf(protoinfo, sizeof(protoinfo), "TCP %s:?? > %s:?? %s %s %s",
srchost, dsthost, tflags, ipinfo, tcpinfo);
} else if (len < (u32) ip->ip_hl * 4 + 16) { // we can get ports an seq
snprintf(tcpinfo, sizeof(tcpinfo), "seq=%lu (incomplete)", (unsigned long) ntohl(tcp->th_seq));
snprintf(protoinfo, sizeof(protoinfo), "TCP %s:%d > %s:%d ?? %s %s",
srchost, ntohs(tcp->th_sport), dsthost, ntohs(tcp->th_dport), ipinfo, tcpinfo);
} else if (len < (u32) ip->ip_hl * 4 + 16 && !frag_off) { // we can't get TCP flags
snprintf(protoinfo, sizeof(protoinfo), "TCP %s:%d > %s:%d ?? %s %s (incomplete)",
srchost, ntohs(tcp->th_sport), dsthost, ntohs(tcp->th_dport),
ipinfo, tcpinfo);
} else { // at least first 16 bytes of TCP header are there (everything we need)
snprintf(tcpinfo, sizeof(tcpinfo), "seq=%lu win=%hi",
(unsigned long) ntohl(tcp->th_seq),
ntohs(tcp->th_win));
p = tflags;
/* These are basically in tcpdump order */
if (tcp->th_flags & TH_SYN) *p++ = 'S';
if (tcp->th_flags & TH_FIN) *p++ = 'F';
if (tcp->th_flags & TH_RST) *p++ = 'R';
if (tcp->th_flags & TH_PUSH) *p++ = 'P';
if (tcp->th_flags & TH_ACK) {
*p++ = 'A';
snprintf(buf, sizeof(buf), " ack=%lu",
(unsigned long) ntohl(tcp->th_ack));
strncat(tcpinfo, buf, sizeof(tcpinfo));
}
if (tcp->th_flags & TH_URG) *p++ = 'U';
if (tcp->th_flags & TH_ECE) *p++ = 'E'; /* rfc 2481/3168 */
if (tcp->th_flags & TH_CWR) *p++ = 'C'; /* rfc 2481/3168 */
*p++ = '\0';
snprintf(protoinfo, sizeof(protoinfo), "TCP %s:%d > %s:%d %s %s %s",
srchost, ntohs(tcp->th_sport), dsthost, ntohs(tcp->th_dport),
tflags, ipinfo, tcpinfo);
}
} else if (ip->ip_p == IPPROTO_UDP && frag_off) {
snprintf(protoinfo, sizeof(protoinfo), "UDP %s:?? > %s:?? fragment %s (incomplete)", srchost, dsthost, ipinfo);
} else if (ip->ip_p == IPPROTO_UDP) {
udp = (udphdr_bsd *) (packet + sizeof(struct ip));
snprintf(protoinfo, sizeof(protoinfo), "UDP %s:%d > %s:%d %s",
srchost, ntohs(udp->uh_sport), dsthost, ntohs(udp->uh_dport),
ipinfo);
} else if (ip->ip_p == IPPROTO_ICMP && frag_off) {
snprintf(protoinfo, sizeof(protoinfo), "ICMP %s > %s fragment %s (incomplete)", srchost, dsthost, ipinfo);
} else if (ip->ip_p == IPPROTO_ICMP) {
char icmptype[128];
struct ppkt {
unsigned char type;
unsigned char code;
unsigned short checksum;
unsigned short id;
unsigned short seq;
} *ping;
ping = (struct ppkt *) ((ip->ip_hl * 4) + (char *) ip);
switch(ping->type) {
case 0:
strcpy(icmptype, "Echo reply"); break;
case 3:
if (ping->code == 0)
strcpy(icmptype, "network unreachable");
else if (ping->code == 1)
strcpy(icmptype, "host unreachable");
else if (ping->code == 2)
strcpy(icmptype, "protocol unreachable");
else if (ping->code == 3)
strcpy(icmptype, "port unreachable");
else if (ping->code == 4)
strcpy(icmptype, "fragmentation required");
else if (ping->code == 5)
strcpy(icmptype, "source route failed");
else if (ping->code == 6)
strcpy(icmptype, "destination network unknown");
else if (ping->code == 7)
strcpy(icmptype, "destination host unknown");
else if (ping->code == 8)
strcpy(icmptype, "source host isolated");
else if (ping->code == 9)
strcpy(icmptype, "destination network administratively prohibited");
else if (ping->code == 10)
strcpy(icmptype, "destination host administratively prohibited");
else if (ping->code == 11)
strcpy(icmptype, "network unreachable for TOS");
else if (ping->code == 12)
strcpy(icmptype, "host unreachable for TOS");
else if (ping->code == 13)
strcpy(icmptype, "communication administratively prohibited by filtering");
else if (ping->code == 14)
strcpy(icmptype, "host precedence violation");
else if (ping->code == 15)
strcpy(icmptype, "precedence cutoff in effect");
else
strcpy(icmptype, "unknown unreachable code");
break;
case 4:
strcpy(icmptype, "source quench"); break;
case 5:
if (ping->code == 0)
strcpy(icmptype, "network redirect");
else if (ping->code == 1)
strcpy(icmptype, "host redirect");
else strcpy(icmptype, "unknown redirect");
break;
case 8:
strcpy(icmptype, "Echo request"); break;
case 11:
if (ping->code == 0)
strcpy(icmptype, "TTL=0 during transit");
else if (ping->code == 1)
strcpy(icmptype, "TTL=0 during reassembly");
else strcpy(icmptype, "TTL exceeded (unknown code)");
break;
case 12:
if (ping->code == 0)
strcpy(icmptype, "IP header bad");
else
strcpy(icmptype, "Misc. parameter problem");
break;
case 13:
strcpy(icmptype, "Timestamp request"); break;
case 14:
strcpy(icmptype, "Timestamp reply"); break;
case 15:
strcpy(icmptype, "Information request"); break;
case 16:
strcpy(icmptype, "Information reply"); break;
case 17:
strcpy(icmptype, "Address mask request"); break;
case 18:
strcpy(icmptype, "Address mask reply"); break;
case 30:
strcpy(icmptype, "Traceroute"); break;
case 37:
strcpy(icmptype, "Domain name request"); break;
case 38:
strcpy(icmptype, "Domain name reply"); break;
case 40:
strcpy(icmptype, "Security failures"); /* rfc 2521 */ break;
default:
strcpy(icmptype, "Unknown type"); break;
break;
}
snprintf(protoinfo, sizeof(protoinfo), "ICMP %s > %s %s (type=%d/code=%d) %s",
srchost, dsthost, icmptype, ping->type, ping->code, ipinfo);
} else {
snprintf(protoinfo, sizeof(protoinfo), "Unknown protocol (%d): %s",
ip->ip_p, ipinfo);
}
return protoinfo;
}
int islocalhost(const struct in_addr * const addr) {
char dev[128];
/* If it is 0.0.0.0 or starts with 127.0.0.1 then it is
probably localhost */
if ((addr->s_addr & htonl(0xFF000000)) == htonl(0x7F000000))
return 1;
if (!addr->s_addr)
return 1;
/* If it is the same addy as a local interface, then it is
probably localhost */
if (ipaddr2devname(dev, addr) != -1)
return 1;
/* OK, so to a first approximation, this addy is probably not
localhost */
return 0;
}
#ifdef WIN32
/* Convert a dnet interface name into the long pcap style. This also caches the data
to speed things up. Fills out pcapdev (up to pcapdevlen) and returns true if it finds anything.
Otherwise returns false. This is only necessary on Windows.*/
bool DnetName2PcapName(const char *dnetdev, char *pcapdev, int pcapdevlen) {
static struct NameCorrelationCache {
char dnetd[64];
char pcapd[128];
} *NCC = NULL;
static int NCCsz = 0;
static int NCCcapacity = 0;
int i;
char tmpdev[128];
// Init the cache if not done yet
if (!NCC) {
NCCcapacity = 5;
NCC = (struct NameCorrelationCache *) safe_zalloc(NCCcapacity * sizeof(*NCC));
NCCsz = 0;
}
// First check if the name is already in the cache
for(i=0; i < NCCsz; i++) {
if (strcmp(NCC[i].dnetd, dnetdev) == 0) {
Strncpy(pcapdev, NCC[i].pcapd, pcapdevlen);
return true;
}
}
// OK, so it isn't in the cache. Let's ask dnet for it.
/* Converts a dnet interface name (ifname) to its pcap equivalent, which is stored in
pcapdev (up to a length of pcapdevlen). Returns 0 and fills in pcapdev if successful. */
if (intf_get_pcap_devname(dnetdev, tmpdev, sizeof(tmpdev)) != 0)
return false;
// We've got it. Let's add it to the cache
if (NCCsz >= NCCcapacity) {
NCCcapacity <<= 2;
NCC = (struct NameCorrelationCache *) safe_realloc(NCC, NCCcapacity * sizeof(*NCC));
}
Strncpy(NCC[NCCsz].dnetd, dnetdev, sizeof(NCC[0].dnetd));
Strncpy(NCC[NCCsz].pcapd, tmpdev, sizeof(NCC[0].pcapd));
NCCsz++;
Strncpy(pcapdev, tmpdev, pcapdevlen);
return true;
}
#endif
pcap_t *my_pcap_open_live(const char *device, int snaplen, int promisc,
int to_ms)
{
char err0r[PCAP_ERRBUF_SIZE];
pcap_t *pt;
char pcapdev[128];
#ifdef WIN32
/* Nmap normally uses device names obtained through dnet for interfaces, but Pcap has its own
naming system. So the conversion is done here */
if (!DnetName2PcapName(device, pcapdev, sizeof(pcapdev))) {
/* Oh crap -- couldn't find the corresponding dev apparently. Let's just go with what we have then ... */
Strncpy(pcapdev, device, sizeof(pcapdev));
}
#else
Strncpy(pcapdev, device, sizeof(pcapdev));
#endif
if (!((pt = pcap_open_live(pcapdev, snaplen, promisc, to_ms, err0r)))) {
fatal("pcap_open_live: %s\nThere are several possible reasons for this, depending on your operating system:\n"
"LINUX: If you are getting Socket type not supported, try modprobe af_packet or recompile your kernel with SOCK_PACKET enabled.\n"
"*BSD: If you are getting device not configured, you need to recompile your kernel with Berkeley Packet Filter support. If you are getting No such file or directory, try creating the device (eg cd /dev; MAKEDEV <device>; or use mknod).\n"
"SOLARIS: If you are trying to scan localhost and getting '/dev/lo0: No such file or directory', complain to Sun. I don't think Solaris can support advanced localhost scans. You can probably use \"-P0 -sT localhost\" though.\n\n", err0r);
}
#ifdef WIN32
/* We want any responses back ASAP */
pcap_setmintocopy(pt, 1);
#endif
return pt;
}
/* Standard BSD internet checksum routine */
unsigned short in_cksum(u16 *ptr,int nbytes) {
register u32 sum;
u16 oddbyte;
register u16 answer;
/*
* Our algorithm is simple, using a 32-bit accumulator (sum),
* we add sequential 16-bit words to it, and at the end, fold back
* all the carry bits from the top 16 bits into the lower 16 bits.
*/
sum = 0;
while (nbytes > 1) {
sum += *ptr++;
nbytes -= 2;
}
/* mop up an odd byte, if necessary */
if (nbytes == 1) {
oddbyte = 0; /* make sure top half is zero */
*((u_char *) &oddbyte) = *(u_char *)ptr; /* one byte only */
sum += oddbyte;
}
/*
* Add back carry outs from top 16 bits to low 16 bits.
*/
sum = (sum >> 16) + (sum & 0xffff); /* add high-16 to low-16 */
sum += (sum >> 16); /* add carry */
answer = ~sum; /* ones-complement, then truncate to 16 bits */
return(answer);
}
/* LEGACY resolve() function that only supports IPv4 -- see IPv6 version
above. Tries to resolve given hostname and stores
result in ip . returns 0 if hostname cannot
be resolved */
int resolve(char *hostname, struct in_addr *ip) {
struct hostent *h;
if (!hostname || !*hostname)
fatal("NULL or zero-length hostname passed to resolve()");
if (inet_aton(hostname, ip))
return 1; /* damn, that was easy ;) */
if ((h = gethostbyname(hostname))) {
memcpy(ip, h->h_addr_list[0], sizeof(struct in_addr));
return 1;
}
return 0;
}
int send_tcp_raw_decoys( int sd, struct eth_nfo *eth,
const struct in_addr *victim, int ttl,
u16 sport, u16 dport, u32 seq, u32 ack, u8 flags,
u16 window, u8 *options, int optlen, char *data,
u16 datalen)
{
int decoy;
for(decoy = 0; decoy < o.numdecoys; decoy++)
if (send_tcp_raw(sd, eth, &o.decoys[decoy], victim, ttl, sport, dport,
seq, ack, flags, window, options, optlen, data,
datalen) == -1)
return -1;
return 0;
}
/* Builds a TCP packet (including an IP header) by packing the fields
with the given information. It allocates a new buffer to store the
packet contents, and then returns that buffer. The packet is not
actually sent by this function. Caller must delete the buffer when
finished with the packet. The packet length is returned in
packetlen, which must be a valid int pointer. */
u8 *build_tcp_raw(const struct in_addr *source,
const struct in_addr *victim, int ttl,
u16 ipid, u16 sport, u16 dport, u32 seq, u32 ack, u8 flags,
u16 window, u8 *options, int optlen, char *data,
u16 datalen, u32 *packetlen) {
struct pseudo_header {
/*for computing TCP checksum, see TCP/IP Illustrated p. 145 */
u32 s_addy;
u32 d_addr;
u8 zer0;
u8 protocol;
u16 length;
};
u8 *packet = (u8 *) safe_malloc(sizeof(struct ip) + sizeof(struct tcphdr) + optlen + datalen);
struct ip *ip = (struct ip *) packet;
struct tcphdr *tcp = (struct tcphdr *) (packet + sizeof(struct ip));
struct pseudo_header *pseudo = (struct pseudo_header *) (packet + sizeof(struct ip) - sizeof(struct pseudo_header));
static int myttl = 0;
assert(victim);
assert(source);
if (optlen % 4) {
fatal("build_tcp_raw() called with an option length argument of %d which is illegal because it is not divisible by 4", optlen);
}
/* Time to live */
if (ttl == -1) {
myttl = (get_random_uint() % 23) + 37;
} else {
myttl = ttl;
}
memset((char *) packet, 0, sizeof(struct ip) + sizeof(struct tcphdr));
pseudo->s_addy = source->s_addr;
pseudo->d_addr = victim->s_addr;
pseudo->protocol = IPPROTO_TCP;
pseudo->length = htons(sizeof(struct tcphdr) + optlen + datalen);
tcp->th_sport = htons(sport);
tcp->th_dport = htons(dport);
if (seq) {
tcp->th_seq = htonl(seq);
}
else if (flags & TH_SYN) {
get_random_bytes(&(tcp->th_seq), 4);
}
if (ack)
tcp->th_ack = htonl(ack);
/*else if (flags & TH_ACK)
tcp->th_ack = rand() + rand();*/
tcp->th_off = 5 + (optlen /4) /*words*/;
tcp->th_flags = flags;
if (window)
tcp->th_win = htons(window);
else tcp->th_win = htons(1024 * (myttl % 4 + 1)); /* Who cares */
/* We should probably copy the data over too */
if (data && datalen)
memcpy(packet + sizeof(struct ip) + sizeof(struct tcphdr) + optlen, data, datalen);
/* And the options */
if (optlen) {
memcpy(packet + sizeof(struct ip) + sizeof(struct tcphdr), options, optlen);
}
#if STUPID_SOLARIS_CHECKSUM_BUG
tcp->th_sum = sizeof(struct tcphdr) + optlen + datalen;
#else
tcp->th_sum = in_cksum((unsigned short *)pseudo, sizeof(struct tcphdr) +
optlen + sizeof(struct pseudo_header) + datalen);
#endif
if ( o.badsum )
--tcp->th_sum;
/* Now for the ip header */
memset(packet, 0, sizeof(struct ip));
ip->ip_v = 4;
ip->ip_hl = 5;
ip->ip_len = htons(sizeof(struct ip) + sizeof(struct tcphdr) + optlen + datalen);
get_random_bytes(&(ip->ip_id), 2);
ip->ip_ttl = myttl;
ip->ip_p = IPPROTO_TCP;
ip->ip_id = ipid;
ip->ip_src.s_addr = source->s_addr;
#ifdef WIN32
// I'm not sure why this is --Fyodor
if (source->s_addr == victim->s_addr) ip->ip_src.s_addr++;
#endif
ip->ip_dst.s_addr= victim->s_addr;
#if HAVE_IP_IP_SUM
ip->ip_sum = in_cksum((unsigned short *)ip, sizeof(struct ip));
#endif
if (TCPIP_DEBUGGING > 1) {
log_write(LOG_STDOUT, "Raw TCP packet creation completed! Here it is:\n");
readtcppacket(packet,ntohs(ip->ip_len));
}
*packetlen = ntohs(ip->ip_len);
return packet;
}
/* You need to call sethdrinclude(sd) on the sending sd before calling this */
int send_tcp_raw( int sd, struct eth_nfo *eth, const struct in_addr *source,
const struct in_addr *victim, int ttl,
u16 sport, u16 dport, u32 seq, u32 ack, u8 flags,
u16 window, u8 *options, int optlen, char *data,
u16 datalen)
{
unsigned int packetlen;
int res = -1;
u8 *packet = build_tcp_raw(source, victim, ttl, get_random_u16(), sport,
dport, seq, ack, flags, window, options, optlen,
data, datalen, &packetlen);
if (!packet) return -1;
res = send_ip_packet(sd, eth, packet, packetlen);
free(packet);
return res;
}
/* Send a pre-built IPv4 packet */
int send_ip_packet(int sd, struct eth_nfo *eth, u8 *packet, unsigned int packetlen) {
struct sockaddr_in sock;
int res;
struct ip *ip = (struct ip *) packet;
struct tcphdr *tcp = NULL;
udphdr_bsd *udp;
u8 *eth_frame;
eth_t *ethsd;
bool ethsd_opened = false;
assert(packet);
assert( (int) packetlen > 0);
// fragmentation requested && packet is bigger than MTU
if (o.fragscan && ( packetlen - ip->ip_hl * 4 > (unsigned int) o.fragscan ))
return send_frag_ip_packet(sd, eth, packet, packetlen, o.fragscan);
if (eth) {
eth_frame = (u8 *) safe_malloc(14 + packetlen);
memcpy(eth_frame + 14, packet, packetlen);
eth_pack_hdr(eth_frame, eth->dstmac, eth->srcmac, ETH_TYPE_IP);
if (!eth->ethsd) {
ethsd = eth_open(eth->devname);
if (!ethsd)
fatal("send_ip_packet: Failed to open ethernet device (%s)", eth->devname);
ethsd_opened = true;
} else ethsd = eth->ethsd;
res = eth_send(ethsd, eth_frame, 14 + packetlen);
PacketTrace::trace(PacketTrace::SENT, packet, packetlen);
if (ethsd_opened)
eth_close(ethsd);
return res;
}
assert(sd >= 0);
memset(&sock, 0, sizeof(sock));
sock.sin_family = AF_INET;
#if HAVE_SOCKADDR_SA_LEN
sock.sin_len = sizeof(sock);
#endif
/* It is bogus that I need the address and port info when sending a RAW IP
packet, but it doesn't seem to work w/o them */
if (packetlen >= 20) {
sock.sin_addr.s_addr = ip->ip_dst.s_addr;
if (ip->ip_p == IPPROTO_TCP && packetlen >= (unsigned int) ip->ip_hl * 4 + 20) {
tcp = (struct tcphdr *) ((u8 *) ip + ip->ip_hl * 4);
sock.sin_port = tcp->th_dport;
} else if (ip->ip_p == IPPROTO_UDP && packetlen >= (unsigned int) ip->ip_hl * 4 + 8) {
udp = (udphdr_bsd *) ((u8 *) ip + ip->ip_hl * 4);
sock.sin_port = udp->uh_dport;
}
}
/* Equally bogus is that the IP total len and IP fragment offset
fields need to be in host byte order on certain BSD variants. I
must deal with it here rather than when building the packet,
because they should be in NBO when I'm sending over raw
ethernet */
#if FREEBSD || BSDI || NETBSD || DEC
ip->ip_len = ntohs(ip->ip_len);
ip->ip_off = ntohs(ip->ip_off);
#endif
res = Sendto("send_ip_packet", sd, packet, packetlen, 0,
(struct sockaddr *)&sock, (int)sizeof(struct sockaddr_in));
return res;
}
/* Create and send all fragments of a pre-built IPv4 packet
* Minimal MTU for IPv4 is 68 and maximal IPv4 header size is 60
* which gives us a right to cut TCP header after 8th byte
* (shouldn't we inflate the header to 60 bytes too?) */
int send_frag_ip_packet(int sd, struct eth_nfo *eth, u8 *packet,
unsigned int packetlen, unsigned int mtu)
{
struct ip *ip = (struct ip *) packet;
int headerlen = ip->ip_hl * 4; // better than sizeof(struct ip)
unsigned int datalen = packetlen - headerlen;
int fdatalen = 0, res = 0;
assert(headerlen <= (int) packetlen);
assert(headerlen >= 20 && headerlen <= 60); // sanity check (RFC791)
assert(mtu > 0 && mtu % 8 == 0); // otherwise, we couldn't set Fragment offset (ip->ip_off) correctly
if (datalen <= mtu) {
error("Warning: fragmentation (mtu=%i) requested but the payload is too small already (%i)", mtu, datalen);
return send_ip_packet(sd, eth, packet, packetlen);
}
u8 *fpacket = (u8 *) safe_malloc(headerlen + mtu);
memcpy(fpacket, packet, headerlen + mtu);
ip = (struct ip *) fpacket;
// create fragments and send them
for (int fragment = 1; fragment * mtu < datalen + mtu; fragment++) {
fdatalen = (fragment * mtu <= datalen ? mtu : datalen % mtu);
ip->ip_len = htons(headerlen + fdatalen);
ip->ip_off = htons((fragment-1) * mtu / 8);
if ((fragment-1) * mtu + fdatalen < datalen)
ip->ip_off |= htons(IP_MF);
#if HAVE_IP_IP_SUM
ip->ip_sum = in_cksum((unsigned short *)ip, headerlen);
#endif
if (fragment > 1) // copy data payload
memcpy(fpacket + headerlen, packet + headerlen + (fragment - 1) * mtu, fdatalen);
res = send_ip_packet(sd, eth, fpacket, headerlen + fdatalen);
if (res == -1)
break;
}
free(fpacket);
return res;
}
/* Builds an ICMP packet (including an IP header) by packing the fields
with the given information. It allocates a new buffer to store the
packet contents, and then returns that buffer. The packet is not
actually sent by this function. Caller must delete the buffer when
finished with the packet. The packet length is returned in
packetlen, which must be a valid int pointer. */
u8 *build_icmp_raw(const struct in_addr *source, const struct in_addr *victim,
int ttl, u16 ipid, u16 seq, unsigned short id, u8 ptype,
u8 pcode, char *data, u16 datalen, u32 *packetlen) {
struct ppkt {
u8 type;
u8 code;
u16 checksum;
u16 id;
u16 seq;
u8 data[1500]; /* Note -- first 4-12 bytes can be used for ICMP header */
} pingpkt;
u32 *datastart = (u32 *) pingpkt.data;
int dlen = sizeof(pingpkt.data);
int icmplen=0;
char *ping = (char *) &pingpkt;
pingpkt.type = ptype;
pingpkt.code = pcode;
if (ptype == 8 && pcode == 0) /* echo request */ {
icmplen = 8;
} else if (ptype == 13 && pcode == 0) /* ICMP timestamp req */ {
icmplen = 20;
memset(datastart, 0, 12);
datastart += 12;
datalen -= 12;
} else if (ptype == 17 && pcode == 0) /* icmp netmask req */ {
icmplen = 12;
*datastart++ = 0;
datalen -= 4;
} else
fatal("Unknown icmp type/code (%d/%d) in build_icmp_raw", ptype, pcode);
if (datalen > 0) {
icmplen += MIN(dlen, datalen);
memset(datastart, 0, MIN(dlen, datalen));
}
/* Fill out the ping packet */
pingpkt.code = 0;
pingpkt.id = id;
pingpkt.seq = seq;
pingpkt.checksum = 0;
pingpkt.checksum = in_cksum((unsigned short *)ping, icmplen);
if ( o.badsum )
--pingpkt.checksum;
return build_ip_raw(source, victim, o.ttl, IPPROTO_ICMP, get_random_u16(),
ping, icmplen, packetlen);
}
void readippacket(const u8 *packet, int readdata) {
struct ip *ip = (struct ip *) packet;
switch(ip->ip_p) {
case IPPROTO_UDP:
readudppacket(packet, readdata);
break;
/* Should add ICMP here at some point */
default:
readtcppacket(packet, readdata);
break;
}
}
/* A simple function I wrote to help in debugging, shows the important fields
of a TCP packet*/
int readtcppacket(const u8 *packet, int readdata) {
struct ip *ip = (struct ip *) packet;
struct tcphdr *tcp = (struct tcphdr *) (packet + sizeof(struct ip));
const unsigned char *data = packet + sizeof(struct ip) + sizeof(struct tcphdr);
int tot_len;
struct in_addr bullshit, bullshit2;
char sourcehost[16];
int i;
int realfrag = 0;
if (!packet) {
fprintf(stderr, "readtcppacket: packet is NULL!\n");
return -1;
}
bullshit.s_addr = ip->ip_src.s_addr; bullshit2.s_addr = ip->ip_dst.s_addr;
/* this is gay */
realfrag = htons(ntohs(ip->ip_off) & 8191 /* 2^13 - 1 */);
tot_len = htons(ip->ip_len);
strncpy(sourcehost, inet_ntoa(bullshit), 16);
i = 4 * (ntohs(ip->ip_hl) + ntohs(tcp->th_off));
if (ip->ip_p== IPPROTO_TCP) {
if (realfrag)
printf("Packet is fragmented, offset field: %u\n", realfrag);
else {
printf("TCP packet: %s:%d -> %s:%d (total: %d bytes)\n", sourcehost,
ntohs(tcp->th_sport), inet_ntoa(bullshit2),
ntohs(tcp->th_dport), tot_len);
printf("Flags: ");
if (!tcp->th_flags) printf("(none)");
if (tcp->th_flags & TH_RST) printf("RST ");
if (tcp->th_flags & TH_SYN) printf("SYN ");
if (tcp->th_flags & TH_ACK) printf("ACK ");
if (tcp->th_flags & TH_PUSH) printf("PSH ");
if (tcp->th_flags & TH_FIN) printf("FIN ");
if (tcp->th_flags & TH_URG) printf("URG ");
printf("\n");
printf("ipid: %hu ttl: %hu ", ntohs(ip->ip_id), ip->ip_ttl);
if (tcp->th_flags & (TH_SYN | TH_ACK)) printf("Seq: %u\tAck: %u\n",
(unsigned int) ntohl(tcp->th_seq), (unsigned int) ntohl(tcp->th_ack));
else if (tcp->th_flags & TH_SYN) printf("Seq: %u\n", (unsigned int) ntohl(tcp->th_seq));
else if (tcp->th_flags & TH_ACK) printf("Ack: %u\n", (unsigned int) ntohl(tcp->th_ack));
}
}
if (readdata && i < tot_len) {
printf("Data portion:\n");
while(i < tot_len) {
printf("%2X%c", data[i], ((i+1) %16)? ' ' : '\n');
i++;
}
printf("\n");
}
return 0;
}
/* A simple function I wrote to help in debugging, shows the important fields
of a UDP packet*/
int readudppacket(const u8 *packet, int readdata) {
struct ip *ip = (struct ip *) packet;
udphdr_bsd *udp = (udphdr_bsd *) (packet + sizeof(struct ip));
const unsigned char *data = packet + sizeof(struct ip) + sizeof(udphdr_bsd);
int tot_len;
struct in_addr bullshit, bullshit2;
char sourcehost[16];
int i;
int realfrag = 0;
if (!packet) {
fprintf(stderr, "readudppacket: packet is NULL!\n");
return -1;
}
bullshit.s_addr = ip->ip_src.s_addr; bullshit2.s_addr = ip->ip_dst.s_addr;
/* this is gay */
realfrag = htons(ntohs(ip->ip_off) & 8191 /* 2^13 - 1 */);
tot_len = htons(ip->ip_len);
strncpy(sourcehost, inet_ntoa(bullshit), 16);
i = 4 * (ntohs(ip->ip_hl)) + 8;
if (ip->ip_p== IPPROTO_UDP) {
if (realfrag)
printf("Packet is fragmented, offset field: %u\n", realfrag);
else {
printf("UDP packet: %s:%d -> %s:%d (total: %d bytes)\n", sourcehost,
ntohs(udp->uh_sport), inet_ntoa(bullshit2),
ntohs(udp->uh_dport), tot_len);
printf("ttl: %hu ", ip->ip_ttl);
}
}
if (readdata && i < tot_len) {
printf("Data portion:\n");
while(i < tot_len) {
printf("%2X%c", data[i], ((i+1)%16)? ' ' : '\n');
i++;
}
printf("\n");
}
return 0;
}
int send_udp_raw_decoys( int sd, struct eth_nfo *eth,
const struct in_addr *victim, int ttl,
u16 sport, u16 dport, u16 ipid, char *data,
u16 datalen) {
int decoy;
for(decoy = 0; decoy < o.numdecoys; decoy++)
if (send_udp_raw(sd, eth, &o.decoys[decoy], victim, ttl, sport, dport, ipid,
data, datalen) == -1)
return -1;
return 0;
}
/* Builds a UDP packet (including an IP header) by packing the fields
with the given information. It allocates a new buffer to store the
packet contents, and then returns that buffer. The packet is not
actually sent by this function. Caller must delete the buffer when
finished with the packet. The packet length is returned in
packetlen, which must be a valid int pointer. */
u8 *build_udp_raw(struct in_addr *source, const struct in_addr *victim,
int ttl, u16 sport, u16 dport, u16 ipid, char *data,
u16 datalen, u32 *packetlen)
{
unsigned char *packet = (unsigned char *) safe_malloc(sizeof(struct ip) + sizeof(udphdr_bsd) + datalen);
struct ip *ip = (struct ip *) packet;
udphdr_bsd *udp = (udphdr_bsd *) (packet + sizeof(struct ip));
static int myttl = 0;
struct pseudo_udp_hdr {
struct in_addr source;
struct in_addr dest;
u8 zero;
u8 proto;
u16 length;
} *pseudo = (struct pseudo_udp_hdr *) ((char *)udp - 12) ;
*packetlen = 0;
/* check that required fields are there and not too silly */
if ( !victim) {
fprintf(stderr, "build_udp_raw: One or more of your parameters suck!\n");
free(packet);
return NULL;
}
/* Time to live */
if (ttl == -1) {
myttl = (get_random_uint() % 23) + 37;
} else {
myttl = ttl;
}
memset((char *) packet, 0, sizeof(struct ip) + sizeof(udphdr_bsd));
udp->uh_sport = htons(sport);
udp->uh_dport = htons(dport);
udp->uh_ulen = htons(8 + datalen);
/* We should probably copy the data over too */
if (data)
memcpy(packet + sizeof(struct ip) + sizeof(udphdr_bsd), data, datalen);
/* Now the pseudo header for checksuming */
pseudo->source.s_addr = source->s_addr;
pseudo->dest.s_addr = victim->s_addr;
pseudo->proto = IPPROTO_UDP;
pseudo->length = htons(sizeof(udphdr_bsd) + datalen);
/* OK, now we should be able to compute a valid checksum */
#if STUPID_SOLARIS_CHECKSUM_BUG
udp->uh_sum = sizeof(udphdr_bsd) + datalen;
#else
udp->uh_sum = in_cksum((unsigned short *)pseudo, 20 /* pseudo + UDP headers */ + datalen);
#endif
if ( o.badsum )
--udp->uh_sum;
/* Goodbye, pseudo header! */
memset(pseudo, 0, sizeof(*pseudo));
/* Now for the ip header */
ip->ip_v = 4;
ip->ip_hl = 5;
ip->ip_len = htons(sizeof(struct ip) + sizeof(udphdr_bsd) + datalen);
ip->ip_id = htons(ipid);
ip->ip_ttl = myttl;
ip->ip_p = IPPROTO_UDP;
ip->ip_src.s_addr = source->s_addr;
#ifdef WIN32
// I'm not exactly sure why this is needed --Fyodor
if(source->s_addr == victim->s_addr) ip->ip_src.s_addr;
#endif
ip->ip_dst.s_addr= victim->s_addr;
#if HAVE_IP_IP_SUM
ip->ip_sum = in_cksum((unsigned short *)ip, sizeof(struct ip));
#endif
if (TCPIP_DEBUGGING > 1) {
printf("Raw UDP packet creation completed! Here it is:\n");
readudppacket(packet,1);
}
*packetlen = ntohs(ip->ip_len);
return packet;
}
int send_udp_raw( int sd, struct eth_nfo *eth, struct in_addr *source,
const struct in_addr *victim,
int ttl, u16 sport, u16 dport, u16 ipid, char *data,
u16 datalen)
{
unsigned int packetlen;
int res = -1;
u8 *packet = build_udp_raw(source, victim, ttl, sport, dport, ipid, data,
datalen, &packetlen);
if (!packet) return -1;
res = send_ip_packet(sd, eth, packet, packetlen);
free(packet);
return res;
}
/* Builds an IP packet (including an IP header) by packing the fields
with the given information. It allocates a new buffer to store the
packet contents, and then returns that buffer. The packet is not
actually sent by this function. Caller must delete the buffer when
finished with the packet. The packet length is returned in
packetlen, which must be a valid int pointer. */
u8 *build_ip_raw(const struct in_addr *source, const struct in_addr *victim,
int ttl, u8 proto, u16 ipid, char *data, u16 datalen,
u32 *packetlen)
{
unsigned char *packet = (unsigned char *) safe_malloc(sizeof(struct ip) + datalen);
struct ip *ip = (struct ip *) packet;
static int myttl = 0;
/* check that required fields are there and not too silly */
if ( !victim) {
fprintf(stderr, "send_ip_raw: One or more of your parameters suck!\n");
free(packet);
return NULL;
}
/* Time to live */
if (ttl == -1) {
myttl = (get_random_uint() % 23) + 37;
} else {
myttl = ttl;
}
memset((char *) packet, 0, sizeof(struct ip));
/* Now for the ip header */
ip->ip_v = 4;
ip->ip_hl = 5;
ip->ip_len = htons(sizeof(struct ip) + datalen);
ip->ip_id = htons(ipid);
ip->ip_ttl = myttl;
ip->ip_p = proto;
ip->ip_src.s_addr = source->s_addr;
// #ifdef WIN32
// TODO: Should this be removed? I'm not sure why this is here -- Fyodor
// if(source->s_addr == victim->s_addr) ip->ip_src.s_addr++;
// #endif
ip->ip_dst.s_addr = victim->s_addr;
#if HAVE_IP_IP_SUM
ip->ip_sum = in_cksum((unsigned short *)ip, sizeof(struct ip));
#endif
/* We should probably copy the data over too */
if (data)
memcpy(packet + sizeof(struct ip), data, datalen);
if (TCPIP_DEBUGGING > 1) {
printf("Raw IP packet creation completed! Here it is:\n");
hdump(packet, ntohs(ip->ip_len));
}
*packetlen = ntohs(ip->ip_len);
return packet;
}
/* You need to call sethdrinclude(sd) on the sending sd before calling this */
int send_ip_raw( int sd, struct eth_nfo *eth, struct in_addr *source,
const struct in_addr *victim, int ttl, u8 proto,
char *data, u16 datalen)
{
unsigned int packetlen;
int res = -1;
u8 *packet = build_ip_raw(source, victim, ttl, proto, get_random_u16(),
data, datalen, &packetlen);
if (!packet) return -1;
res = send_ip_packet(sd, eth, packet, packetlen);
free(packet);
return res;
}
int unblock_socket(int sd) {
#ifdef WIN32
u_long one = 1;
if(sd != 501) // Hack related to WinIP Raw Socket support
ioctlsocket (sd, FIONBIO, &one);
#else
int options;
/*Unblock our socket to prevent recvfrom from blocking forever
on certain target ports. */
options = O_NONBLOCK | fcntl(sd, F_GETFL);
fcntl(sd, F_SETFL, options);
#endif //WIN32
return 1;
}
/* Read an IP packet using libpcap . We return the packet and take
a pcap descripter and a pointer to the packet length (which we set
in the function. If you want a maximum length returned, you
should specify that in pcap_open_live() */
/* to_usec is the timeout period in microseconds -- use 0 to skip the
test and -1 to block forever. Note that we don't interrupt pcap, so
low values (and 0) degenerate to the timeout specified
in pcap_open_live()
*/
/* If rcvdtime is non-null and a packet is returned, rcvd will be
filled with the time that packet was captured from the wire by
pcap. If linknfo is not NULL, linknfo->headerlen and
linknfo->header will be filled with the appropriate values. */
char *readip_pcap(pcap_t *pd, unsigned int *len, long to_usec,
struct timeval *rcvdtime, struct link_header *linknfo) {
unsigned int offset = 0;
struct pcap_pkthdr head;
char *p;
int datalink;
int timedout = 0;
struct timeval tv_start, tv_end;
static char *alignedbuf = NULL;
static unsigned int alignedbufsz=0;
static int warning = 0;
if (linknfo) { memset(linknfo, 0, sizeof(*linknfo)); }
if (!pd) fatal("NULL packet device passed to readip_pcap");
if (to_usec < 0) {
if (!warning) {
warning = 1;
error("WARNING: Negative timeout value (%lu) passed to readip_pcap() -- using 0", to_usec);
}
to_usec = 0;
}
/* New packet capture device, need to recompute offset */
if ( (datalink = pcap_datalink(pd)) < 0)
fatal("Cannot obtain datalink information: %s", pcap_geterr(pd));
/* NOTE: IF A NEW OFFSET EVER EXCEEDS THE CURRENT MAX (24), ADJUST
MAX_LINK_HEADERSZ in tcpip.h */
switch(datalink) {
case DLT_EN10MB: offset = 14; break;
case DLT_IEEE802: offset = 22; break;
#ifdef __amigaos__
case DLT_MIAMI: offset = 16; break;
#endif
#ifdef DLT_LOOP
case DLT_LOOP:
#endif
case DLT_NULL: offset = 4; break;
case DLT_SLIP:
#ifdef DLT_SLIP_BSDOS
case DLT_SLIP_BSDOS:
#endif
#if (FREEBSD || OPENBSD || NETBSD || BSDI || MACOSX)
offset = 16;
#else
offset = 24; /* Anyone use this??? */
#endif
break;
case DLT_PPP:
#ifdef DLT_PPP_BSDOS
case DLT_PPP_BSDOS:
#endif
#ifdef DLT_PPP_SERIAL
case DLT_PPP_SERIAL:
#endif
#ifdef DLT_PPP_ETHER
case DLT_PPP_ETHER:
#endif
#if (FREEBSD || OPENBSD || NETBSD || BSDI || MACOSX)
offset = 4;
#else
#ifdef SOLARIS
offset = 8;
#else
offset = 24; /* Anyone use this? */
#endif /* ifdef solaris */
#endif /* if freebsd || openbsd || netbsd || bsdi */
break;
case DLT_RAW: offset = 0; break;
case DLT_FDDI: offset = 21; break;
#ifdef DLT_ENC
case DLT_ENC: offset = 12; break;
#endif /* DLT_ENC */
#ifdef DLT_LINUX_SLL
case DLT_LINUX_SLL: offset = 16; break;
#endif
default:
p = (char *) pcap_next(pd, &head);
if (head.caplen == 0) {
/* Lets sleep a brief time and try again to increase the chance of seeing
a real packet ... */
usleep(500000);
p = (char *) pcap_next(pd, &head);
}
if (head.caplen > 100000) {
fatal("FATAL: readip_pcap: bogus caplen from libpcap (%d) on interface type %d", head.caplen, datalink);
}
error("FATAL: Unknown datalink type (%d). Caplen: %d; Packet:\n", datalink, head.caplen);
lamont_hdump(p, head.caplen);
exit(1);
}
if (to_usec > 0) {
gettimeofday(&tv_start, NULL);
}
do {
#ifdef WIN32
gettimeofday(&tv_end, NULL);
long to_left = MAX(1, (to_usec - TIMEVAL_SUBTRACT(tv_end, tv_start)) / 1000);
// Set the timeout (BUGBUG: this is cheating)
PacketSetReadTimeout(pd->adapter, to_left);
#endif
p = (char *) pcap_next(pd, &head);
if (p) {
if (head.caplen <= offset) {
*len = 0;
return NULL;
}
if (offset && linknfo) {
linknfo->datalinktype = datalink;
linknfo->headerlen = offset;
assert(offset < MAX_LINK_HEADERSZ);
memcpy(linknfo->header, p, MIN(sizeof(linknfo->header), offset));
}
p += offset;
}
if (!p || (*p & 0x40) != 0x40) {
/* Should we timeout? */
if (to_usec == 0) {
timedout = 1;
} else if (to_usec > 0) {
gettimeofday(&tv_end, NULL);
if (TIMEVAL_SUBTRACT(tv_end, tv_start) >= to_usec) {
timedout = 1;
}
}
}
} while(!timedout && (!p || (*p & 0x40) != 0x40)); /* Go until we get IPv4 packet */
if (timedout) {
*len = 0;
return NULL;
}
*len = head.caplen - offset;
if (*len > alignedbufsz) {
alignedbuf = (char *) realloc(alignedbuf, *len);
if (!alignedbuf) {
fatal("Unable to realloc %d bytes of mem", *len);
}
alignedbufsz = *len;
}
memcpy(alignedbuf, p, *len);
// printf("Just got a packet at %li,%li\n", head.ts.tv_sec, head.ts.tv_usec);
if (rcvdtime) {
// FIXME: I eventually need to figure out why Windows head.ts time is sometimes BEFORE the time I
// sent the packet (which is according to gettimeofday() in nbase). For now, I will sadly have to
// use gettimeofday() for Windows in this case
// Actually I now allow .05 discrepancy. So maybe this isn't needed. I'll comment out for now.
// Nope: it is still needed at least for Windows. Sometimes the time from he pcap header is a
// COUPLE SECONDS before the gettimeofday() results :(.
#if defined(WIN32) || defined(__amigaos__)
gettimeofday(&tv_end, NULL);
*rcvdtime = tv_end;
#else
rcvdtime->tv_sec = head.ts.tv_sec;
rcvdtime->tv_usec = head.ts.tv_usec;
assert(head.ts.tv_sec);
#endif
}
if (rcvdtime)
PacketTrace::trace(PacketTrace::RCVD, (u8 *) alignedbuf, *len, rcvdtime);
else PacketTrace::trace(PacketTrace::RCVD, (u8 *) alignedbuf, *len);
return alignedbuf;
}
// Returns whether the packet receive time value obtained from libpcap
// (and thus by readip_pcap()) should be considered valid. When
// invalid (Windows and Amiga), readip_pcap returns the time you called it.
bool pcap_recv_timeval_valid() {
#if defined(WIN32) || defined(__amigaos__)
return false;
#else
return true;
#endif
}
/* A trivial functon that maintains a cache of IP to MAC Address
entries. If the command is ARPCACHE_GET, this func looks for the
IPv4 address in ss and fills in the 'mac' parameter and returns
true if it is found. Otherwise (not found), the function returns
false. If the command is ARPCACHE_SET, the function adds an entry
with the given ip (ss) and mac address. An existing entry for the
IP ss will be overwritten with the new MAC address. true is always
returned for the set command. */
bool NmapArpCache(int command, struct sockaddr_storage *ss, u8 *mac) {
struct sockaddr_in *sin = (struct sockaddr_in *) ss;
struct ArpCache {
u32 ip; /* Network byte order */
u8 mac[6];
};
static struct ArpCache *Cache = NULL;
static int ArpCapacity = 0;
static int ArpCacheSz = 0;
int i;
if (sin->sin_family != AF_INET)
fatal("NmapArpCache() can only take IPv4 addresses. Sorry");
if (command == ARPCACHE_GET) {
for(i=0; i < ArpCacheSz; i++) {
if (Cache[i].ip == sin->sin_addr.s_addr) {
memcpy(mac, Cache[i].mac, 6);
return true;
}
}
return false;
}
assert(command == ARPCACHE_SET);
if (ArpCacheSz == ArpCapacity) {
if (ArpCapacity == 0) ArpCapacity = 32;
else ArpCapacity <<= 2;
}
Cache = (struct ArpCache *) safe_realloc(Cache,
ArpCapacity * sizeof(struct ArpCache));
/* Ensure that it isn't already there ... */
for(i=0; i < ArpCacheSz; i++) {
if (Cache[i].ip == sin->sin_addr.s_addr) {
memcpy(Cache[i].mac, mac, 6);
return true;
}
}
/* Add it to the end of the list */
Cache[i].ip = sin->sin_addr.s_addr;
memcpy(Cache[i].mac, mac, 6);
return true;
}
/* Attempts to read one IPv4/Ethernet ARP reply packet from the pcap
descriptor pd. If it receives one, fills in sendermac (must pass
in 6 bytes), senderIP, and rcvdtime (can be NULL if you don't care)
and returns 1. If it times out and reads no arp requests, returns
0. to_usec is the timeout period in microseconds. Use 0 to avoid
blocking to the extent possible. Returns
-1 or exits if ther is an error. */
int read_arp_reply_pcap(pcap_t *pd, u8 *sendermac, struct in_addr *senderIP,
long to_usec, struct timeval *rcvdtime) {
static int warning = 0;
int datalink;
struct pcap_pkthdr head;
u8 *p;
int timedout = 0;
int badcounter = 0;
struct timeval tv_start, tv_end;
if (!pd) fatal("NULL packet device passed to readarp_reply_pcap");
if (to_usec < 0) {
if (!warning) {
warning = 1;
error("WARNING: Negative timeout value (%lu) passed to %s() -- using 0", to_usec, __FUNCTION__);
}
to_usec = 0;
}
/* New packet capture device, need to recompute offset */
if ( (datalink = pcap_datalink(pd)) < 0)
fatal("Cannot obtain datalink information: %s", pcap_geterr(pd));
if (datalink != DLT_EN10MB)
fatal("readarp_reply_pcap called on interfaces that is datatype %d rather than DLT_EN10MB (%d)", datalink, DLT_EN10MB);
if (to_usec > 0) {
gettimeofday(&tv_start, NULL);
}
do {
#ifdef WIN32
if (to_usec == 0)
PacketSetReadTimeout(pd->adapter, 1);
else {
gettimeofday(&tv_end, NULL);
long to_left = MAX(1, (to_usec - TIMEVAL_SUBTRACT(tv_end, tv_start)) / 1000);
// Set the timeout (BUGBUG: this is cheating)
PacketSetReadTimeout(pd->adapter, to_left);
}
#endif
p = (u8 *) pcap_next(pd, &head);
if (p && head.caplen >= 42) { /* >= because Ethernet padding makes 60 */
/* frame type 0x0806 (arp), hw type eth (0x0001), prot ip (0x0800),
hw size (0x06), prot size (0x04) */
if (memcmp(p + 12, "\x08\x06\x00\x01\x08\x00\x06\x04\x00\x02", 10) == 0) {
memcpy(sendermac, p + 22, 6);
/* I think alignment should allow this ... */
memcpy(&senderIP->s_addr, p+28, 4);
break;
}
}
if (!p) {
/* Should we timeout? */
if (to_usec == 0) {
timedout = 1;
} else if (to_usec > 0) {
gettimeofday(&tv_end, NULL);
if (TIMEVAL_SUBTRACT(tv_end, tv_start) >= to_usec) {
timedout = 1;
}
}
} else {
/* We'll be a bit patient if we're getting actual packets back, but
not indefinitely so */
if (badcounter++ > 50)
timedout = 1;
}
} while(!timedout);
if (timedout) return 0;
if (rcvdtime) {
// FIXME: I eventually need to figure out why Windows head.ts time is sometimes BEFORE the time I
// sent the packet (which is according to gettimeofday() in nbase). For now, I will sadly have to
// use gettimeofday() for Windows in this case
// Actually I now allow .05 discrepancy. So maybe this isn't needed. I'll comment out for now.
// Nope: it is still needed at least for Windows. Sometimes the time from he pcap header is a
// COUPLE SECONDS before the gettimeofday() results :(.
#if defined(WIN32) || defined(__amigaos__)
gettimeofday(&tv_end, NULL);
*rcvdtime = tv_end;
#else
rcvdtime->tv_sec = head.ts.tv_sec;
rcvdtime->tv_usec = head.ts.tv_usec;
assert(head.ts.tv_sec);
#endif
}
PacketTrace::traceArp(PacketTrace::RCVD, (u8 *) p, 42, rcvdtime);
return 1;
}
/* This function tries to determine the target's ethernet MAC address
from a received packet as follows:
1) If linkhdr is an ethernet header, grab the src mac (otherwise give up)
2) If overwrite is 0 and a MAC is already set for this target, give up.
3) If the packet source address is not the target, give up.
4) Use the routing table to try to determine rather target is
directly connected to the src host running Nmap. If it is, set the MAC.
This function returns 0 if it ends up setting the MAC, nonzero otherwise
This function assumes that ip has already been verified as
containing a complete IP header (or at least the first 20 bytes).
*/
int setTargetMACIfAvailable(Target *target, struct link_header *linkhdr,
struct ip *ip, int overwrite) {
if (!linkhdr || !target || !ip)
return 1;
if (linkhdr->datalinktype != DLT_EN10MB || linkhdr->headerlen != 14)
return 2;
if (!overwrite && target->MACAddress())
return 3;
if (ip->ip_src.s_addr != target->v4host().s_addr)
return 4;
/* Sometimes bogus MAC address still gets through, like during some localhost scans */
if (memcmp(linkhdr->header+6, "\0\0\0\0\0\0", 6) == 0)
return 5;
if (target->ifType() == devt_ethernet && target->directlyConnected()) {
/* Yay! This MAC address seems valid */
target->setMACAddress(linkhdr->header + 6);
return 0;
}
return 5;
}
/* Issues an ARP request for the MAC of targetss (which will be placed
in targetmac if obtained) from the source IP (srcip) and source mac
(srcmac) given. "The request is ussued using device dev to the
broadcast MAC address. The transmission is attempted up to 3
times. If none of these elicit a response, false will be returned.
If the mac is determined, true is returned. */
bool doArp(const char *dev, const u8 *srcmac,
const struct sockaddr_storage *srcip,
const struct sockaddr_storage *targetip, u8 *targetmac) {
/* timeouts in microseconds ... the first ones are retransmit times, while
the final one is when we give up */
int timeouts[] = { 100000, 400000, 800000 };
int max_sends = 3;
int num_sends = 0; // How many we have sent so far
eth_t *ethsd;
u8 frame[ETH_HDR_LEN + ARP_HDR_LEN + ARP_ETHIP_LEN];
const struct sockaddr_in *targetsin = (struct sockaddr_in *) targetip;
const struct sockaddr_in *srcsin = (struct sockaddr_in *) srcip;
struct timeval start, now, rcvdtime;
int timeleft;
int listenrounds;
int rc;
pcap_t *pd;
struct in_addr rcvdIP;
bool foundit = false;
if (targetsin->sin_family != AF_INET || srcsin->sin_family != AF_INET)
fatal("%s can only handle IPv4 addresses", __FUNCTION__);
/* Start listening */
pd = my_pcap_open_live(dev, 50, 1, 25);
set_pcap_filter(dev, pd, "arp and ether dst host %02X:%02X:%02X:%02X:%02X:%02X", srcmac[0], srcmac[1], srcmac[2], srcmac[3], srcmac[4], srcmac[5]);
/* Prepare probe and sending stuff */
ethsd = eth_open(dev);
if (!ethsd) fatal("%s: failed to open device %s", __FUNCTION__, dev);
eth_pack_hdr(frame, ETH_ADDR_BROADCAST, *srcmac, ETH_TYPE_ARP);
arp_pack_hdr_ethip(frame + ETH_HDR_LEN, ARP_OP_REQUEST, *srcmac,
srcsin->sin_addr, ETH_ADDR_BROADCAST,
targetsin->sin_addr);
gettimeofday(&start, NULL);
gettimeofday(&now, NULL);
while(!foundit && num_sends < max_sends) {
/* Send the sucker */
rc = eth_send(ethsd, frame, sizeof(frame));
if (rc != sizeof(frame)) {
error("WARNING: %s: eth_send of ARP packet returned %u rather than expected %d bytes\n", __FUNCTION__, rc, (int) sizeof(frame));
}
PacketTrace::traceArp(PacketTrace::SENT, (u8 *) frame, sizeof(frame), &now);
num_sends++;
listenrounds = 0;
while(!foundit) {
gettimeofday(&now, NULL);
timeleft = timeouts[num_sends - 1] - TIMEVAL_SUBTRACT(now, start);
if (timeleft < 0) {
if (listenrounds > 0) break;
else timeleft = 25000;
}
listenrounds++;
/* Now listen until we reach our next timeout or get an answer */
rc = read_arp_reply_pcap(pd, targetmac, &rcvdIP, timeleft, &rcvdtime);
if (rc == -1) fatal("%s: Received -1 response from readarp_reply_pcap",
__FUNCTION__);
if (rc == 1) {
/* Yay, I got one! But is it the right one? */
if (rcvdIP.s_addr != targetsin->sin_addr.s_addr)
continue; /* D'oh! */
foundit = true; /* WOOHOO! */
}
}
}
/* OK - let's close up shop ... */
pcap_close(pd);
eth_close(ethsd);
return foundit;
}
/* This function ensures that the next hop MAC address for a target is
filled in. This address is the target's own MAC if it is directly
connected, and the next hop mac otherwise. Returns true if the
address is set when the function ends, false if not. This function
firt checks if it is already set, if not it tries the arp cache,
and if that fails it sends an ARP request itself. This should be
called after an ARP scan if many directly connected machines are
involved. setDirectlyConnected() (whether true or false) should
have already been called on target before this. The target device
and src mac address should also already be set. */
bool setTargetNextHopMAC(Target *target) {
struct sockaddr_storage targetss, srcss;
size_t sslen;
arp_t *a;
u8 mac[6];
struct arp_entry ae;
if (target->ifType() != devt_ethernet)
return false; /* Duh. */
/* First check if we already have it, duh. */
if (target->NextHopMACAddress())
return true;
/* For connected machines, it is the same as the target addy */
if (target->directlyConnected() && target->MACAddress()) {
target->setNextHopMACAddress(target->MACAddress());
return true;
}
if (target->directlyConnected()) {
target->TargetSockAddr(&targetss, &sslen);
} else {
if (!target->nextHop(&targetss, &sslen))
fatal("%s: Failed to determine nextHop to target", __FUNCTION__);
}
/* First, let us check the Nmap arp cache ... */
if (NmapArpCache(ARPCACHE_GET, &targetss, mac)) {
target->setNextHopMACAddress(mac);
return true;
}
/* Maybe the system ARP cache will be more helpful */
a = arp_open();
addr_ston((sockaddr *)&targetss, &ae.arp_pa);
if (arp_get(a, &ae) == 0) {
NmapArpCache(ARPCACHE_SET, &targetss, ae.arp_ha.addr_eth.data);
target->setNextHopMACAddress(ae.arp_ha.addr_eth.data);
return true;
}
/* OK, the last choice is to send our own damn ARP request (and
retransmissions if necessary) to determine the MAC */
target->SourceSockAddr(&srcss, NULL);
if (doArp(target->deviceName(), target->SrcMACAddress(), &srcss, &targetss,
mac)) {
NmapArpCache(ARPCACHE_SET, &targetss, mac);
target->setNextHopMACAddress(mac);
return true;
}
/* I'm afraid that we couldn't find it! Maybe it doesn't exist?*/
return false;
}
/* Set a pcap filter */
void set_pcap_filter(const char *device,
pcap_t *pd, char *bpf, ...)
{
va_list ap;
char buf[3072];
struct bpf_program fcode;
#ifndef __amigaos__
unsigned int localnet, netmask;
#else
bpf_u_int32 localnet, netmask;
#endif
char err0r[256];
// Cast below is becaue OpenBSD apparently has a version that takes a
// non-const device (hopefully they don't actually write to it).
if (pcap_lookupnet( (char *) device, &localnet, &netmask, err0r) < 0)
fatal("Failed to lookup subnet/netmask for device (%s): %s", device, err0r);
va_start(ap, bpf);
if (vsnprintf(buf, sizeof(buf), bpf, ap) >= (int) sizeof(buf))
fatal("set_pcap_filter called with too-large filter arg\n");
va_end(ap);
/* Due to apparent bug in libpcap */
/* Maybe this bug no longer exists ... I'll comment out for now
* if (islocalhost(target->v4hostip()))
* buf[0] = '\0'; */
if (o.debugging)
log_write(LOG_STDOUT, "Packet capture filter (device %s): %s\n", device, buf);
if (pcap_compile(pd, &fcode, buf, 0, netmask) < 0)
fatal("Error compiling our pcap filter: %s\n", pcap_geterr(pd));
if (pcap_setfilter(pd, &fcode) < 0 )
fatal("Failed to set the pcap filter: %s\n", pcap_geterr(pd));
}
/* The 'dev' passed in must be at least 32 bytes long */
int ipaddr2devname( char *dev, const struct in_addr *addr ) {
struct interface_info *mydevs;
int numdevs;
int i;
struct sockaddr_in *sin;
mydevs = getinterfaces(&numdevs);
if (!mydevs) return -1;
for(i=0; i < numdevs; i++) {
sin = (struct sockaddr_in *) &mydevs[i].addr;
if (sin->sin_family != AF_INET)
continue;
if (addr->s_addr == sin->sin_addr.s_addr) {
Strncpy(dev, mydevs[i].devname, 32);
return 0;
}
}
return -1;
}
int devname2ipaddr(char *dev, struct in_addr *addr) {
struct interface_info *mydevs;
int numdevs;
int i;
mydevs = getinterfaces(&numdevs);
if (!mydevs) return -1;
for(i=0; i < numdevs; i++) {
if (!strcmp(dev, mydevs[i].devfullname)) {
memcpy(addr, (char *) &mydevs[i].addr, sizeof(struct in_addr));
return 0;
}
}
return -1;
}
struct dnet_collector_route_nfo {
struct sys_route *routes;
int numroutes;
int capacity; /* Capacity of routes or ifaces, depending on context */
struct interface_info *ifaces;
int numifaces;
};
int collect_dnet_routes(const struct route_entry *entry, void *arg) {
struct dnet_collector_route_nfo *dcrn = (struct dnet_collector_route_nfo *) arg;
int i;
/* Make sure that it is the proper type of route ... */
if (entry->route_dst.addr_type != ADDR_TYPE_IP || entry->route_gw.addr_type != ADDR_TYPE_IP)
return 0; /* Not interested in IPv6 routes at the moment ... */
/* Make sure we have room for the new route */
if (dcrn->numroutes >= dcrn->capacity) {
dcrn->capacity <<= 2;
dcrn->routes = (struct sys_route *) realloc(dcrn->routes,
dcrn->capacity * sizeof(struct sys_route));
}
/* Now for the important business */
dcrn->routes[dcrn->numroutes].dest = entry->route_dst.addr_ip;
addr_btom(entry->route_dst.addr_bits, &dcrn->routes[dcrn->numroutes].netmask, sizeof(dcrn->routes[dcrn->numroutes].netmask));
dcrn->routes[dcrn->numroutes].gw.s_addr = entry->route_gw.addr_ip;
/* Now determine which interface the route relates to */
u32 mask;
struct sockaddr_in *sin;
for(i = 0; i < dcrn->numifaces; i++) {
sin = (struct sockaddr_in *) &dcrn->ifaces[i].addr;
mask = htonl((unsigned long) (0-1) << (32 - dcrn->ifaces[i].netmask_bits));
if ((sin->sin_addr.s_addr & mask) == (entry->route_gw.addr_ip & mask)) {
dcrn->routes[dcrn->numroutes].device = &dcrn->ifaces[i];
break;
}
}
if (i == dcrn->numifaces) {
error("WARNING: Unable to find appropriate interface for system route to %s\n", addr_ntoa(&entry->route_gw));
return 0;
}
dcrn->numroutes++;
return 0;
}
int collect_dnet_interfaces(const struct intf_entry *entry, void *arg) {
struct dnet_collector_route_nfo *dcrn = (struct dnet_collector_route_nfo *) arg;
int i;
int numifaces = dcrn->numifaces;
/* Make sure we have room for the new route */
if (dcrn->numifaces >= dcrn->capacity) {
dcrn->capacity <<= 2;
dcrn->ifaces = (struct interface_info *) realloc(dcrn->ifaces,
dcrn->capacity * sizeof(struct interface_info));
}
if (entry->intf_addr.addr_type == ADDR_TYPE_IP) {
addr_ntos(&entry->intf_addr, (struct sockaddr *) &dcrn->ifaces[numifaces].addr);
dcrn->ifaces[numifaces].netmask_bits = entry->intf_addr.addr_bits;
} else {
for(i=0; i < (int) entry->intf_alias_num; i++) {
if (entry->intf_alias_addrs[i].addr_type == ADDR_TYPE_IP) {
addr_ntos(&entry->intf_alias_addrs[i], (struct sockaddr *) &dcrn->ifaces[numifaces].addr);
dcrn->ifaces[numifaces].netmask_bits = entry->intf_alias_addrs[i].addr_bits;
break;
}
if (i == (int) entry->intf_alias_num)
return 0; /* No IPv4 addresses found for this interface */
}
}
/* OK, address/netmask found. Let's get the name */
Strncpy(dcrn->ifaces[numifaces].devname, entry->intf_name, sizeof(dcrn->ifaces[numifaces].devname));
Strncpy(dcrn->ifaces[numifaces].devfullname, entry->intf_name, sizeof(dcrn->ifaces[numifaces].devfullname));
/* Interface type */
if (entry->intf_type == INTF_TYPE_ETH) {
dcrn->ifaces[numifaces].device_type = devt_ethernet;
/* Collect the MAC address since this is ethernet */
memcpy(dcrn->ifaces[numifaces].mac, &entry->intf_link_addr.addr_eth.data, 6);
}
else if (entry->intf_type == INTF_TYPE_LOOPBACK)
dcrn->ifaces[numifaces].device_type = devt_loopback;
else if (entry->intf_type == INTF_TYPE_TUN)
dcrn->ifaces[numifaces].device_type = devt_p2p;
else dcrn->ifaces[numifaces].device_type = devt_other;
/* Is the interface up and running? */
dcrn->ifaces[numifaces].device_up = (entry->intf_flags & INTF_FLAG_UP)? true : false;
/* For the rest of the information, we must open the interface directly ... */
dcrn->numifaces++;
return 0;
}
struct interface_info *getinterfaces(int *howmany) {
static bool initialized = 0;
static struct interface_info *mydevs;
static int numifaces = 0;
int ii_capacity = 0;
#if WIN32
struct dnet_collector_route_nfo dcrn;
intf_t *it;
#else //!WIN32
int sd;
struct ifconf ifc;
struct ifreq *ifr;
struct ifreq tmpifr;
#endif
int len, rc;
char *p;
u8 *buf;
int bufsz;
struct sockaddr_in *sin;
u16 ifflags;
if (!initialized) {
initialized = 1;
ii_capacity = 16;
mydevs = (struct interface_info *) safe_zalloc(sizeof(struct interface_info) * ii_capacity);
#if WIN32
/* On Win32 we just use Dnet to determine the interface list */
dcrn.routes = NULL;
dcrn.numroutes = 0;
dcrn.capacity = ii_capacity; // I'm reusing this struct for ii now
dcrn.ifaces = mydevs;
dcrn.numifaces = 0;
it = intf_open();
if (!it) fatal("%s: intf_open() failed", __FUNCTION__);
if (intf_loop(it, collect_dnet_interfaces, &dcrn) != 0)
fatal("%s: intf_loop() failed", __FUNCTION__);
intf_close(it);
mydevs = dcrn.ifaces;
numifaces = dcrn.numifaces;
ii_capacity = dcrn.capacity;
#else // !Win32
/* Dummy socket for ioctl */
sd = socket(AF_INET, SOCK_DGRAM, 0);
if (sd < 0) pfatal("socket in getinterfaces");
bufsz = 20480;
buf = (u8 *) safe_zalloc(bufsz);
ifc.ifc_len = bufsz;
ifc.ifc_buf = (char *) buf;
if (ioctl(sd, SIOCGIFCONF, &ifc) < 0) {
fatal("Failed to determine your configured interfaces!\n");
}
ifr = (struct ifreq *) buf;
if (ifc.ifc_len == 0)
fatal("getinterfaces: SIOCGIFCONF claims you have no network interfaces!\n");
#if HAVE_SOCKADDR_SA_LEN
/* len = MAX(sizeof(struct sockaddr), ifr->ifr_addr.sa_len);*/
len = ifr->ifr_addr.sa_len + sizeof(ifr->ifr_name);
#else
len = sizeof(struct ifreq);
/* len = sizeof(SA); */
#endif
#if TCPIP_DEBUGGING
printf("ifnet list length = %d\n",ifc.ifc_len);
printf("sa_len = %d\n",len);
hdump((unsigned char *) buf, ifc.ifc_len);
printf("ifr = %X\n",(unsigned)(*(char **)&ifr));
printf("Size of struct ifreq: %d\n", sizeof(struct ifreq));
#endif
for(; ifr && ifr->ifr_name[0] && ((u8 *)ifr) < buf + ifc.ifc_len;
ifr = (struct ifreq *)(((char *)ifr) + len)) {
#if TCPIP_DEBUGGING
printf("ifr_name size = %d\n", sizeof(ifr->ifr_name));
printf("ifr = %X\n",(unsigned)(*(char **)&ifr));
#endif
/* On some platforms (such as FreeBSD), the length of each ifr changes
based on the sockaddr type used, so we get the next length now */
#if HAVE_SOCKADDR_SA_LEN
len = ifr->ifr_addr.sa_len + sizeof(ifr->ifr_name);
#endif
/* skip any device with no name */
if (!*((char *)ifr))
continue;
/* We currently only handle IPv4 */
sin = (struct sockaddr_in *) &ifr->ifr_addr;
if (sin->sin_family != AF_INET)
continue;
memcpy(&(mydevs[numifaces].addr), sin, MIN(sizeof(mydevs[numifaces].addr), sizeof(*sin)));
Strncpy(mydevs[numifaces].devname, ifr->ifr_name, sizeof(mydevs[numifaces].devname));
/* devname isn't allowed to have alias qualification */
if ((p = strchr(mydevs[numifaces].devname, ':')))
*p = '\0';
Strncpy(mydevs[numifaces].devfullname, ifr->ifr_name, sizeof(mydevs[numifaces].devfullname));
Strncpy(tmpifr.ifr_name, ifr->ifr_name, sizeof(tmpifr.ifr_name));
memcpy(&(tmpifr.ifr_addr), sin, MIN(sizeof(tmpifr.ifr_addr), sizeof(*sin)));
rc = ioctl(sd, SIOCGIFNETMASK, &tmpifr);
if (rc < 0 && errno != EADDRNOTAVAIL)
pfatal("Failed to determine the netmask of %s!", tmpifr.ifr_name);
else if (rc < 0)
mydevs[numifaces].netmask_bits = 32;
else {
sin = (struct sockaddr_in *) &(tmpifr.ifr_addr); /* ifr_netmask only on Linux */
addr_stob(&(tmpifr.ifr_addr), &mydevs[numifaces].netmask_bits);
}
// printf("ifr name=%s addr=%s, mask=%X\n", mydevs[numifaces].name, inet_ntoa(mydevs[numifaces].addr), mydevs[numifaces].netmask.s_addr);
/* Now we need to determine the device type ... this technique
is kinda iffy ... may not be portable. */
/* First we get the flags */
Strncpy(tmpifr.ifr_name, ifr->ifr_name, sizeof(tmpifr.ifr_name));
memcpy(&(tmpifr.ifr_addr), sin, MIN(sizeof(tmpifr.ifr_addr), sizeof(*sin)));
rc = ioctl(sd, SIOCGIFFLAGS, &tmpifr);
if (rc < 0) fatal("Failed to get IF Flags for device %s", ifr->ifr_name);
ifflags = tmpifr.ifr_flags;
if (ifflags & IFF_LOOPBACK)
mydevs[numifaces].device_type = devt_loopback;
else if (ifflags & IFF_BROADCAST) {
mydevs[numifaces].device_type = devt_ethernet;
/* Get the MAC Address ... */
#ifdef SIOCGIFHWADDR
Strncpy(tmpifr.ifr_name, mydevs[numifaces].devname, sizeof(tmpifr.ifr_name));
memcpy(&(tmpifr.ifr_addr), sin, MIN(sizeof(tmpifr.ifr_addr), MIN(sizeof(tmpifr.ifr_addr), sizeof(*sin))));
rc = ioctl(sd, SIOCGIFHWADDR, &tmpifr);
if (rc < 0 && errno != EADDRNOTAVAIL)
pfatal("Failed to determine the MAC address of %s!", tmpifr.ifr_name);
else if (rc >= 0)
memcpy(mydevs[numifaces].mac, &tmpifr.ifr_addr.sa_data, 6);
#else
/* Let's just let libdnet handle it ... */
eth_t *ethsd = eth_open(mydevs[numifaces].devname);
eth_addr_t ethaddr;
if (!ethsd)
fatal("%s: Failed to open ethernet interface (%s). A possible cause on BSD operating systems is running out of BPF devices (see http://seclists.org/lists/nmap-dev/2006/Jan-Mar/0014.html).", __FUNCTION__,
mydevs[numifaces].devname);
if (eth_get(ethsd, &ethaddr) != 0)
fatal("%s: Failed to obtain MAC address for ethernet interface (%s)",
__FUNCTION__, mydevs[numifaces].devname);
memcpy(mydevs[numifaces].mac, ethaddr.data, 6);
#endif /*SIOCGIFHWADDR*/
}
else if (ifflags & IFF_POINTOPOINT)
mydevs[numifaces].device_type = devt_p2p;
else mydevs[numifaces].device_type = devt_other;
if (ifflags & IFF_UP)
mydevs[numifaces].device_up = true;
else mydevs[numifaces].device_up = false;
numifaces++;
if (numifaces == ii_capacity) {
ii_capacity <<= 2;
mydevs = (struct interface_info *) realloc(mydevs, sizeof(struct interface_info) * ii_capacity);
assert(mydevs);
}
mydevs[numifaces].devname[0] = mydevs[numifaces].devfullname[0] = '\0';
}
free(buf);
close(sd);
#endif //!WIN32
}
if (howmany) *howmany = numifaces;
return mydevs;
}
/* Looks for an interface assigned to the given IP (ss), and returns
the interface_info for the first one found. If non found, returns NULL */
struct interface_info *getInterfaceByIP(struct sockaddr_storage *ss) {
struct sockaddr_in *sin = (struct sockaddr_in *) ss;
struct sockaddr_in *ifsin;
struct interface_info *ifaces;
int numifaces = 0;
int ifnum;
if (sin->sin_family != AF_INET)
fatal("%s called with non-IPv4 address", __FUNCTION__);
ifaces = getinterfaces(&numifaces);
for(ifnum=0; ifnum < numifaces; ifnum++) {
ifsin = (struct sockaddr_in *) &ifaces[ifnum].addr;
if (ifsin->sin_family != AF_INET) continue;
if (sin->sin_addr.s_addr == ifsin->sin_addr.s_addr)
return &ifaces[ifnum];
}
return NULL;
}
/* Looks for an interface with the given name (iname), and returns the
corresponding interface_info if found. Will accept a match of
devname or devfullname. Returns NULL if none found */
struct interface_info *getInterfaceByName(char *iname) {
struct interface_info *ifaces;
int numifaces = 0;
int ifnum;
ifaces = getinterfaces(&numifaces);
for(ifnum=0; ifnum < numifaces; ifnum++) {
if (strcmp(ifaces[ifnum].devfullname, iname) == 0 ||
strcmp(ifaces[ifnum].devname, iname) == 0)
return &ifaces[ifnum];
}
return NULL;
}
/* A trivial function used with qsort to sort the routes by netmask */
static int nmaskcmp(const void *a, const void *b) {
struct sys_route *r1 = (struct sys_route *) a;
struct sys_route *r2 = (struct sys_route *) b;
if (r1->netmask == r2->netmask)
return 0;
if (ntohl(r1->netmask) > ntohl(r2->netmask))
return -1;
else return 1;
}
/* Parse the system routing table, converting each route into a
sys_route entry. Returns an array of sys_routes. numroutes is set
to the number of routes in the array. The routing table is only
read the first time this is called -- later results are cached.
The returned route array is sorted by netmask with the most
specific matches first. */
struct sys_route *getsysroutes(int *howmany) {
int route_capacity = 128;
static struct sys_route *routes = NULL;
static int numroutes = 0;
FILE *routefp;
char buf[1024];
char iface[16];
char *p, *endptr;
struct interface_info *ifaces;
int numifaces = 0;
int i;
u32 mask;
struct sockaddr_in *sin;
struct interface_info *ii;
if (!routes) {
routes = (struct sys_route *) safe_zalloc(route_capacity * sizeof(struct sys_route));
ifaces = getinterfaces(&numifaces);
/* First let us try Linux-style /proc/net/route */
routefp = fopen("/proc/net/route", "r");
if (routefp) {
fgets(buf, sizeof(buf), routefp); /* Kill the first line (column headers) */
while(fgets(buf,sizeof(buf), routefp)) {
p = strtok(buf, " \t\n");
if (!p) {
error("Could not find interface in /proc/net/route line");
continue;
}
if (*p == '*')
continue; /* Deleted route -- any other valid reason for
a route to start with an asterict? */
Strncpy(iface, p, sizeof(iface));
p = strtok(NULL, " \t\n");
endptr = NULL;
routes[numroutes].dest = strtoul(p, &endptr, 16);
if (!endptr || *endptr) {
error("Failed to determine Destination from /proc/net/route");
continue;
}
/* Now for the gateway */
p = strtok(NULL, " \t\n");
if (!p) break;
endptr = NULL;
routes[numroutes].gw.s_addr = strtoul(p, &endptr, 16);
if (!endptr || *endptr) {
error("Failed to determine gw for %s from /proc/net/route\n", iface);
}
for(i=0; i < 5; i++) {
p = strtok(NULL, " \t\n");
if (!p) break;
}
if (!p) {
error("Failed to find field %d in /proc/net/route", i + 2);
continue;
}
endptr = NULL;
routes[numroutes].netmask = strtoul(p, &endptr, 16);
if (!endptr || *endptr) {
error("Failed to determine mask from /proc/net/route");
continue;
}
for(i=0; i < numifaces; i++) {
if (!strcmp(iface, ifaces[i].devfullname)) {
routes[numroutes].device = &ifaces[i];
break;
}
}
if (i == numifaces) {
error("Failed to find device %s which was referenced in /proc/net/route", iface);
continue;
}
/* Now to deal with some alias nonsense ... at least on Linux
this file will just list the short name, even though IP
information (such as source address) from an alias must be
used. So if the purported device can't reach the gateway,
try to find a device that starts with the same short
devname, but can (e.g. eth0 -> eth0:3) */
ii = &ifaces[i];
mask = htonl((unsigned long) (0-1) << (32 - ii->netmask_bits));
sin = (struct sockaddr_in *) &ii->addr;
if (routes[numroutes].gw.s_addr && (sin->sin_addr.s_addr & mask) !=
(routes[numroutes].gw.s_addr & mask)) {
for(i=0; i < numifaces; i++) {
if (ii == &ifaces[i]) continue;
if (strcmp(ii->devname, ifaces[i].devname) == 0) {
sin = (struct sockaddr_in *) &ifaces[i].addr;
if ((sin->sin_addr.s_addr & mask) ==
(routes[numroutes].gw.s_addr & mask)) {
routes[numroutes].device = &ifaces[i];
}
}
}
}
numroutes++;
if (numroutes >= route_capacity) {
route_capacity <<= 2;
routes = (struct sys_route *) realloc(routes, route_capacity * sizeof(struct sys_route));
}
}
} else {
struct dnet_collector_route_nfo dcrn;
dcrn.routes = routes;
dcrn.numroutes = numroutes;
dcrn.capacity = route_capacity;
dcrn.ifaces = ifaces;
dcrn.numifaces = numifaces;
route_t *dr = route_open();
if (!dr) fatal("%s: route_open() failed", __FUNCTION__);
if (route_loop(dr, collect_dnet_routes, &dcrn) != 0) {
fatal("%s: route_loop() failed", __FUNCTION__);
}
route_close(dr);
/* These values could have changed in the callback */
route_capacity = dcrn.capacity;
numroutes = dcrn.numroutes;
routes = dcrn.routes;
}
/* Ensure that the route array is sorted by netmask */
for(i=1; i < numroutes; i++) {
if (ntohl(routes[i].netmask) > ntohl(routes[i-1].netmask))
break;
}
if (i < numroutes) {
/* they're not sorted ... better take care of that */
qsort(routes, numroutes, sizeof(routes[0]), nmaskcmp);
}
}
if (!howmany) fatal("NULL howmany ptr passed to getsysroutes()");
*howmany = numroutes;
return routes;
}
/* Takes a destination address (dst) and tries to determine the
source address and interface necessary to route to this address.
If no route is found, false is returned and rnfo is undefined. If
a route is found, true is returned and rnfo is filled in with all
of the routing details. This function takes into account -S and -e
options set by user (o.spoofsource, o.device) */
bool route_dst(const struct sockaddr_storage *const dst, struct route_nfo *rnfo) {
struct interface_info *ifaces;
struct interface_info *iface = NULL;
int numifaces = 0;
struct sys_route *routes;
struct sockaddr_storage spoofss;
size_t spoofsslen;
int numroutes = 0;
int ifnum;
int i;
u32 mask;
struct sockaddr_in *ifsin, *dstsin;
if (!dst) fatal("route_nfo passed a NULL dst address");
dstsin = (struct sockaddr_in *)dst;
if (dstsin->sin_family != AF_INET)
fatal("Sorry -- route_dst currently only supports IPv4");
/* First let us deal with the case where a user requested a specific spoofed IP/dev */
if (o.spoofsource || *o.device) {
if (o.spoofsource) {
o.SourceSockAddr(&spoofss, &spoofsslen);
if (!*o.device) {
/* Look up the device corresponding to src IP, if any ... */
iface = getInterfaceByIP(&spoofss);
}
}
if (*o.device) {
iface = getInterfaceByName(o.device);
if (!iface)
fatal("Could not find interface %s which was specified by -e", o.device);
}
if (iface) {
/* Is it directly connected? */
mask = htonl((unsigned long) (0-1) << (32 - iface->netmask_bits));
ifsin = (struct sockaddr_in *) &(iface->addr);
if ((ifsin->sin_addr.s_addr & mask) == (dstsin->sin_addr.s_addr & mask))
rnfo->direct_connect = 1;
else {
rnfo->direct_connect = 0;
/* must find the next hop by checking route table ... */
routes = getsysroutes(&numroutes);
/* Now we simply go through the list and take the first match */
for(i=0; i < numroutes; i++) {
if (strcmp(routes[i].device->devname, iface->devname) == 0 &&
(routes[i].dest & routes[i].netmask) ==
(dstsin->sin_addr.s_addr & routes[i].netmask)) {
/* Yay, found a matching route. */
ifsin = (struct sockaddr_in *) &rnfo->nexthop;
ifsin->sin_family = AF_INET;
ifsin->sin_addr.s_addr = routes[i].gw.s_addr;
}
}
}
memcpy(&rnfo->ii, iface, sizeof(rnfo->ii));
if (o.spoofsource)
memcpy(&rnfo->srcaddr, &spoofss, sizeof(rnfo->srcaddr));
else
memcpy(&rnfo->srcaddr, &(iface->addr), sizeof(rnfo->srcaddr));
return true;
}
/* Control will get here if -S was specified to a non-interface
IP, but no interface was specified with -e. We will try to
determine the proper interface in that case */
}
ifaces = getinterfaces(&numifaces);
/* I suppose that I'll first determine whether it is a direct connect instance */
for(ifnum=0; ifnum < numifaces; ifnum++) {
ifsin = (struct sockaddr_in *) &ifaces[ifnum].addr;
if (ifsin->sin_family != AF_INET) continue;
if (dstsin->sin_addr.s_addr == ifsin->sin_addr.s_addr &&
ifaces[ifnum].device_type != devt_loopback) {
/* Trying to scan one of the machine's own interfaces -- we need
to use the localhost device for this */
for(i=0; i < numifaces; i++)
if (ifaces[i].device_type == devt_loopback)
break;
if (i < numifaces) {
rnfo->direct_connect = true;
memcpy(&rnfo->ii, &ifaces[i], sizeof(rnfo->ii));
/* But the source address we want to use is the target addy */
if (o.spoofsource)
memcpy(&rnfo->srcaddr, &spoofss, sizeof(rnfo->srcaddr));
else
memcpy(&rnfo->srcaddr, &ifaces[ifnum].addr, sizeof(rnfo->srcaddr));
return true;
}
/* Hmmm ... no localhost -- I guess I'll just try using the device
itself */
}
mask = htonl((unsigned long) (0-1) << (32 - ifaces[ifnum].netmask_bits));
if ((ifsin->sin_addr.s_addr & mask) == (dstsin->sin_addr.s_addr & mask)) {
rnfo->direct_connect = 1;
memcpy(&rnfo->ii, &ifaces[ifnum], sizeof(rnfo->ii));
if (o.spoofsource)
memcpy(&rnfo->srcaddr, &spoofss, sizeof(rnfo->srcaddr));
else
memcpy(&rnfo->srcaddr, &ifaces[ifnum].addr, sizeof(rnfo->srcaddr));
return true;
}
}
/* OK, so it isn't directly connected. Let's do some routing! */
rnfo->direct_connect = false;
routes = getsysroutes(&numroutes);
/* Now we simply go through the list and take the first match */
for(i=0; i < numroutes; i++) {
if ((routes[i].dest & routes[i].netmask) ==
(dstsin->sin_addr.s_addr & routes[i].netmask)) {
/* Yay, found a matching route. */
memcpy(&rnfo->ii, routes[i].device, sizeof(rnfo->ii));
if (o.spoofsource)
memcpy(&rnfo->srcaddr, &spoofss, sizeof(rnfo->srcaddr));
else
memcpy(&rnfo->srcaddr, &routes[i].device->addr, sizeof(rnfo->srcaddr));
ifsin = (struct sockaddr_in *) &rnfo->nexthop;
ifsin->sin_family = AF_INET;
ifsin->sin_addr.s_addr = routes[i].gw.s_addr;
return true;
}
}
return false;
}
/* Maximize the receive buffer of a socket descriptor (up to 500K) */
void max_rcvbuf(int sd) {
int optval = 524288 /*2^19*/;
recvfrom6_t optlen = sizeof(int);
#ifndef WIN32
if (setsockopt(sd, SOL_SOCKET, SO_RCVBUF, (const char *) &optval, optlen))
if (o.debugging) perror("Problem setting large socket receive buffer");
if (o.debugging) {
getsockopt(sd, SOL_SOCKET, SO_RCVBUF,(char *) &optval, &optlen);
log_write(LOG_STDOUT, "Our buffer size is now %d\n", optval);
}
#endif /* WIN32 */
}
/* Maximize the open file descriptor limit for this process go up to the
max allowed */
int max_sd() {
#ifndef WIN32
struct rlimit r;
static int maxfds = -1;
if (maxfds > 0)
return maxfds;
#if(defined(RLIMIT_NOFILE))
if (!getrlimit(RLIMIT_NOFILE, &r)) {
r.rlim_cur = r.rlim_max;
if (setrlimit(RLIMIT_NOFILE, &r))
if (o.debugging) perror("setrlimit RLIMIT_NOFILE failed");
if (!getrlimit(RLIMIT_NOFILE, &r)) {
maxfds = r.rlim_cur;
return maxfds;
} else return 0;
}
#endif
#if(defined(RLIMIT_OFILE) && !defined(RLIMIT_NOFILE))
if (!getrlimit(RLIMIT_OFILE, &r)) {
r.rlim_cur = r.rlim_max;
if (setrlimit(RLIMIT_OFILE, &r))
if (o.debugging) perror("setrlimit RLIMIT_OFILE failed");
if (!getrlimit(RLIMIT_OFILE, &r)) {
maxfds = r.rlim_cur;
return maxfds;
}
else return 0;
}
#endif
#endif /* WIN32 */
return 0;
}
/* Convert a socket to blocking mode */
int block_socket(int sd) {
#ifdef WIN32
unsigned long options=0;
if(sd == 501) return 1;
ioctlsocket(sd, FIONBIO, (unsigned long *)&options);
#else
int options;
options = (~O_NONBLOCK) & fcntl(sd, F_GETFL);
fcntl(sd, F_SETFL, options);
#endif
return 1;
}
/* Give broadcast permission to a socket */
void broadcast_socket(int sd) {
int one = 1;
#ifdef WIN32
if(sd == 501) return;
#endif
if (setsockopt(sd, SOL_SOCKET, SO_BROADCAST, (const char *)&one, sizeof(int)) != 0) {
fprintf(stderr, "Failed to secure socket broadcasting permission\n");
perror("setsockopt");
}
}
/* Do a receive (recv()) on a socket and stick the results (up to
len) into buf . Give up after 'seconds'. Returns the number of
bytes read (or -1 in the case of an error. It only does one recv
(it will not keep going until len bytes are read). If timedout is
not NULL, it will be set to zero (no timeout occured) or 1 (it
did). */
int recvtime(int sd, char *buf, int len, int seconds, int *timedout) {
int res;
struct timeval timeout;
fd_set readfd;
timeout.tv_sec = seconds;
timeout.tv_usec = 0;
FD_ZERO(&readfd);
FD_SET(sd, &readfd);
if (timedout) *timedout = 0;
res = select(sd + 1, &readfd, NULL, NULL, &timeout);
if (res > 0 ) {
res = recv(sd, buf, len, 0);
if (res >= 0) return res;
perror("recv in recvtime");
return 0;
}
else if (!res) {
if (timedout) *timedout = 1;
return 0;
}
perror("select() in recvtime");
return -1;
}
/* Examines the given tcp packet and obtains the TCP timestamp option
information if available. Note that the CALLER must ensure that
"tcp" contains a valid header (in particular the th_off must be the
true packet length and tcp must contain it). If a valid timestamp
option is found in the header, nonzero is returned and the
'timestamp' and 'echots' parameters are filled in with the
appropriate value (if non-null). Otherwise 0 is returned and the
parameters (if non-null) are filled with 0. Remember that the
correct way to check for errors is to look at the return value
since a zero ts or echots could possibly be valid. */
int gettcpopt_ts(struct tcphdr *tcp, u32 *timestamp, u32 *echots) {
unsigned char *p;
int len = 0;
int op;
int oplen;
/* first we find where the tcp options start ... */
p = ((unsigned char *)tcp) + 20;
len = 4 * tcp->th_off - 20;
while(len > 0 && *p != 0 /* TCPOPT_EOL */) {
op = *p++;
if (op == 0 /* TCPOPT_EOL */) break;
if (op == 1 /* TCPOPT_NOP */) { len--; continue; }
oplen = *p++;
if (oplen < 2) break; /* No infinite loops, please */
if (oplen > len) break; /* Not enough space */
if (op == 8 /* TCPOPT_TIMESTAMP */ && oplen == 10) {
/* Legitimate ts option */
if (timestamp) {
memcpy((char *) timestamp, p, 4);
*timestamp = ntohl(*timestamp);
}
p += 4;
if (echots) {
memcpy((char *) echots, p, 4);
*echots = ntohl(*echots);
}
return 1;
}
len -= oplen;
p += oplen - 2;
}
/* Didn't find anything */
if (timestamp) *timestamp = 0;
if (echots) *echots = 0;
return 0;
}
int Sendto(char *functionname, int sd, const unsigned char *packet, int len,
unsigned int flags, struct sockaddr *to, int tolen) {
struct sockaddr_in *sin = (struct sockaddr_in *) to;
int res;
int retries = 0;
int sleeptime = 0;
do {
if ((res = sendto(sd, (const char *) packet, len, flags, to, tolen)) == -1) {
int err = socket_errno();
error("sendto in %s: sendto(%d, packet, %d, 0, %s, %d) => %s",
functionname, sd, len, inet_ntoa(sin->sin_addr), tolen,
strerror(err));
#if WIN32
return -1;
#else
if (retries > 2 || err == EPERM || err == EACCES || err == EADDRNOTAVAIL
|| err == EINVAL)
return -1;
sleeptime = 15 * (1 << (2 * retries));
error("Sleeping %d seconds then retrying", sleeptime);
fflush(stderr);
sleep(sleeptime);
#endif
}
retries++;
} while( res == -1);
PacketTrace::trace(PacketTrace::SENT, packet, len);
return res;
}
IPProbe::IPProbe() {
packetbuflen = 0;
packetbuf = NULL;
Reset();
}
void IPProbe::Reset() {
if (packetbuf)
free(packetbuf);
packetbuflen = 0;
packetbuf = NULL;
ipv4 = NULL;
icmp = NULL;
tcp = NULL;
udp = NULL;
}
IPProbe::~IPProbe() {
if (packetbuf) {
free(packetbuf);
packetbuf = NULL;
packetbuflen = 0;
}
Reset();
}
int IPProbe::storePacket(u8 *ippacket, u32 len) {
assert(packetbuf == NULL);
af = AF_INET;
packetbuf = (u8 *) safe_malloc(len);
memcpy(packetbuf, ippacket, len);
packetbuflen = len;
ipv4 = (struct ip *) packetbuf;
assert(ipv4->ip_v == 4);
assert(len >= 20);
assert(len == (u32) ntohs(ipv4->ip_len));
if (ipv4->ip_p == IPPROTO_TCP) {
if (len >= (unsigned) ipv4->ip_hl * 4 + 20)
tcp = (struct tcphdr *) ((u8 *) ipv4 + ipv4->ip_hl * 4);
} else if (ipv4->ip_p == IPPROTO_ICMP) {
if (len >= (unsigned) ipv4->ip_hl * 4 + 8)
icmp = (struct icmp *) ((u8 *) ipv4 + ipv4->ip_hl * 4);
} else if (ipv4->ip_p == IPPROTO_UDP) {
if (len >= (unsigned) ipv4->ip_hl * 4 + 8)
udp = (udphdr_bsd *) ((u8 *) ipv4 + ipv4->ip_hl * 4);
}
return 0;
}
ArpProbe::ArpProbe() {
packetbuflen = 0;
packetbuf = NULL;
Reset();
}
void ArpProbe::Reset() {
if (packetbuf)
free(packetbuf);
packetbuflen = 0;
packetbuf = NULL;
ipquery = NULL;
}
ArpProbe::~ArpProbe() {
if (packetbuf) {
free(packetbuf);
packetbuf = NULL;
packetbuflen = 0;
}
Reset();
}
int ArpProbe::storePacket(u8 *arppacket, u32 len) {
assert(packetbuf == NULL);
assert(len == 42);
packetbuf = (u8 *) safe_malloc(len);
memcpy(packetbuf, arppacket, len);
packetbuflen = len;
ipquery = (struct in_addr *) ((u8 *)arppacket + 38);
return 0;
}