freebsd-skq/sbin/ipfw/ipfw2.c
2010-03-04 16:55:32 +00:00

3890 lines
90 KiB
C

/*
* Copyright (c) 2002-2003 Luigi Rizzo
* Copyright (c) 1996 Alex Nash, Paul Traina, Poul-Henning Kamp
* Copyright (c) 1994 Ugen J.S.Antsilevich
*
* Idea and grammar partially left from:
* Copyright (c) 1993 Daniel Boulet
*
* Redistribution and use in source forms, with and without modification,
* are permitted provided that this entire comment appears intact.
*
* Redistribution in binary form may occur without any restrictions.
* Obviously, it would be nice if you gave credit where credit is due
* but requiring it would be too onerous.
*
* This software is provided ``AS IS'' without any warranties of any kind.
*
* NEW command line interface for IP firewall facility
*
* $FreeBSD$
*/
#include <sys/types.h>
#include <sys/socket.h>
#include <sys/sockio.h>
#include <sys/sysctl.h>
#include "ipfw2.h"
#include <ctype.h>
#include <err.h>
#include <errno.h>
#include <grp.h>
#include <netdb.h>
#include <pwd.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sysexits.h>
#include <time.h> /* ctime */
#include <timeconv.h> /* _long_to_time */
#include <unistd.h>
#include <fcntl.h>
#include <net/ethernet.h>
#include <net/if.h> /* only IFNAMSIZ */
#include <netinet/in.h>
#include <netinet/in_systm.h> /* only n_short, n_long */
#include <netinet/ip.h>
#include <netinet/ip_icmp.h>
#include <netinet/ip_fw.h>
#include <netinet/tcp.h>
#include <arpa/inet.h>
struct cmdline_opts co; /* global options */
int resvd_set_number = RESVD_SET;
#define GET_UINT_ARG(arg, min, max, tok, s_x) do { \
if (!av[0]) \
errx(EX_USAGE, "%s: missing argument", match_value(s_x, tok)); \
if (_substrcmp(*av, "tablearg") == 0) { \
arg = IP_FW_TABLEARG; \
break; \
} \
\
{ \
long _xval; \
char *end; \
\
_xval = strtol(*av, &end, 10); \
\
if (!isdigit(**av) || *end != '\0' || (_xval == 0 && errno == EINVAL)) \
errx(EX_DATAERR, "%s: invalid argument: %s", \
match_value(s_x, tok), *av); \
\
if (errno == ERANGE || _xval < min || _xval > max) \
errx(EX_DATAERR, "%s: argument is out of range (%u..%u): %s", \
match_value(s_x, tok), min, max, *av); \
\
if (_xval == IP_FW_TABLEARG) \
errx(EX_DATAERR, "%s: illegal argument value: %s", \
match_value(s_x, tok), *av); \
arg = _xval; \
} \
} while (0)
static void
PRINT_UINT_ARG(const char *str, uint32_t arg)
{
if (str != NULL)
printf("%s",str);
if (arg == IP_FW_TABLEARG)
printf("tablearg");
else
printf("%u", arg);
}
static struct _s_x f_tcpflags[] = {
{ "syn", TH_SYN },
{ "fin", TH_FIN },
{ "ack", TH_ACK },
{ "psh", TH_PUSH },
{ "rst", TH_RST },
{ "urg", TH_URG },
{ "tcp flag", 0 },
{ NULL, 0 }
};
static struct _s_x f_tcpopts[] = {
{ "mss", IP_FW_TCPOPT_MSS },
{ "maxseg", IP_FW_TCPOPT_MSS },
{ "window", IP_FW_TCPOPT_WINDOW },
{ "sack", IP_FW_TCPOPT_SACK },
{ "ts", IP_FW_TCPOPT_TS },
{ "timestamp", IP_FW_TCPOPT_TS },
{ "cc", IP_FW_TCPOPT_CC },
{ "tcp option", 0 },
{ NULL, 0 }
};
/*
* IP options span the range 0 to 255 so we need to remap them
* (though in fact only the low 5 bits are significant).
*/
static struct _s_x f_ipopts[] = {
{ "ssrr", IP_FW_IPOPT_SSRR},
{ "lsrr", IP_FW_IPOPT_LSRR},
{ "rr", IP_FW_IPOPT_RR},
{ "ts", IP_FW_IPOPT_TS},
{ "ip option", 0 },
{ NULL, 0 }
};
static struct _s_x f_iptos[] = {
{ "lowdelay", IPTOS_LOWDELAY},
{ "throughput", IPTOS_THROUGHPUT},
{ "reliability", IPTOS_RELIABILITY},
{ "mincost", IPTOS_MINCOST},
{ "congestion", IPTOS_ECN_CE},
{ "ecntransport", IPTOS_ECN_ECT0},
{ "ip tos option", 0},
{ NULL, 0 }
};
static struct _s_x limit_masks[] = {
{"all", DYN_SRC_ADDR|DYN_SRC_PORT|DYN_DST_ADDR|DYN_DST_PORT},
{"src-addr", DYN_SRC_ADDR},
{"src-port", DYN_SRC_PORT},
{"dst-addr", DYN_DST_ADDR},
{"dst-port", DYN_DST_PORT},
{NULL, 0}
};
/*
* we use IPPROTO_ETHERTYPE as a fake protocol id to call the print routines
* This is only used in this code.
*/
#define IPPROTO_ETHERTYPE 0x1000
static struct _s_x ether_types[] = {
/*
* Note, we cannot use "-:&/" in the names because they are field
* separators in the type specifications. Also, we use s = NULL as
* end-delimiter, because a type of 0 can be legal.
*/
{ "ip", 0x0800 },
{ "ipv4", 0x0800 },
{ "ipv6", 0x86dd },
{ "arp", 0x0806 },
{ "rarp", 0x8035 },
{ "vlan", 0x8100 },
{ "loop", 0x9000 },
{ "trail", 0x1000 },
{ "at", 0x809b },
{ "atalk", 0x809b },
{ "aarp", 0x80f3 },
{ "pppoe_disc", 0x8863 },
{ "pppoe_sess", 0x8864 },
{ "ipx_8022", 0x00E0 },
{ "ipx_8023", 0x0000 },
{ "ipx_ii", 0x8137 },
{ "ipx_snap", 0x8137 },
{ "ipx", 0x8137 },
{ "ns", 0x0600 },
{ NULL, 0 }
};
static struct _s_x rule_actions[] = {
{ "accept", TOK_ACCEPT },
{ "pass", TOK_ACCEPT },
{ "allow", TOK_ACCEPT },
{ "permit", TOK_ACCEPT },
{ "count", TOK_COUNT },
{ "pipe", TOK_PIPE },
{ "queue", TOK_QUEUE },
{ "divert", TOK_DIVERT },
{ "tee", TOK_TEE },
{ "netgraph", TOK_NETGRAPH },
{ "ngtee", TOK_NGTEE },
{ "fwd", TOK_FORWARD },
{ "forward", TOK_FORWARD },
{ "skipto", TOK_SKIPTO },
{ "deny", TOK_DENY },
{ "drop", TOK_DENY },
{ "reject", TOK_REJECT },
{ "reset6", TOK_RESET6 },
{ "reset", TOK_RESET },
{ "unreach6", TOK_UNREACH6 },
{ "unreach", TOK_UNREACH },
{ "check-state", TOK_CHECKSTATE },
{ "//", TOK_COMMENT },
{ "nat", TOK_NAT },
{ "reass", TOK_REASS },
{ "setfib", TOK_SETFIB },
{ NULL, 0 } /* terminator */
};
static struct _s_x rule_action_params[] = {
{ "altq", TOK_ALTQ },
{ "log", TOK_LOG },
{ "tag", TOK_TAG },
{ "untag", TOK_UNTAG },
{ NULL, 0 } /* terminator */
};
/*
* The 'lookup' instruction accepts one of the following arguments.
* -1 is a terminator for the list.
* Arguments are passed as v[1] in O_DST_LOOKUP options.
*/
static int lookup_key[] = {
TOK_DSTIP, TOK_SRCIP, TOK_DSTPORT, TOK_SRCPORT,
TOK_UID, TOK_JAIL, -1 };
static struct _s_x rule_options[] = {
{ "tagged", TOK_TAGGED },
{ "uid", TOK_UID },
{ "gid", TOK_GID },
{ "jail", TOK_JAIL },
{ "in", TOK_IN },
{ "limit", TOK_LIMIT },
{ "keep-state", TOK_KEEPSTATE },
{ "bridged", TOK_LAYER2 },
{ "layer2", TOK_LAYER2 },
{ "out", TOK_OUT },
{ "diverted", TOK_DIVERTED },
{ "diverted-loopback", TOK_DIVERTEDLOOPBACK },
{ "diverted-output", TOK_DIVERTEDOUTPUT },
{ "xmit", TOK_XMIT },
{ "recv", TOK_RECV },
{ "via", TOK_VIA },
{ "fragment", TOK_FRAG },
{ "frag", TOK_FRAG },
{ "fib", TOK_FIB },
{ "ipoptions", TOK_IPOPTS },
{ "ipopts", TOK_IPOPTS },
{ "iplen", TOK_IPLEN },
{ "ipid", TOK_IPID },
{ "ipprecedence", TOK_IPPRECEDENCE },
{ "iptos", TOK_IPTOS },
{ "ipttl", TOK_IPTTL },
{ "ipversion", TOK_IPVER },
{ "ipver", TOK_IPVER },
{ "estab", TOK_ESTAB },
{ "established", TOK_ESTAB },
{ "setup", TOK_SETUP },
{ "tcpdatalen", TOK_TCPDATALEN },
{ "tcpflags", TOK_TCPFLAGS },
{ "tcpflgs", TOK_TCPFLAGS },
{ "tcpoptions", TOK_TCPOPTS },
{ "tcpopts", TOK_TCPOPTS },
{ "tcpseq", TOK_TCPSEQ },
{ "tcpack", TOK_TCPACK },
{ "tcpwin", TOK_TCPWIN },
{ "icmptype", TOK_ICMPTYPES },
{ "icmptypes", TOK_ICMPTYPES },
{ "dst-ip", TOK_DSTIP },
{ "src-ip", TOK_SRCIP },
{ "dst-port", TOK_DSTPORT },
{ "src-port", TOK_SRCPORT },
{ "proto", TOK_PROTO },
{ "MAC", TOK_MAC },
{ "mac", TOK_MAC },
{ "mac-type", TOK_MACTYPE },
{ "verrevpath", TOK_VERREVPATH },
{ "versrcreach", TOK_VERSRCREACH },
{ "antispoof", TOK_ANTISPOOF },
{ "ipsec", TOK_IPSEC },
{ "icmp6type", TOK_ICMP6TYPES },
{ "icmp6types", TOK_ICMP6TYPES },
{ "ext6hdr", TOK_EXT6HDR},
{ "flow-id", TOK_FLOWID},
{ "ipv6", TOK_IPV6},
{ "ip6", TOK_IPV6},
{ "ipv4", TOK_IPV4},
{ "ip4", TOK_IPV4},
{ "dst-ipv6", TOK_DSTIP6},
{ "dst-ip6", TOK_DSTIP6},
{ "src-ipv6", TOK_SRCIP6},
{ "src-ip6", TOK_SRCIP6},
{ "lookup", TOK_LOOKUP},
{ "//", TOK_COMMENT },
{ "not", TOK_NOT }, /* pseudo option */
{ "!", /* escape ? */ TOK_NOT }, /* pseudo option */
{ "or", TOK_OR }, /* pseudo option */
{ "|", /* escape */ TOK_OR }, /* pseudo option */
{ "{", TOK_STARTBRACE }, /* pseudo option */
{ "(", TOK_STARTBRACE }, /* pseudo option */
{ "}", TOK_ENDBRACE }, /* pseudo option */
{ ")", TOK_ENDBRACE }, /* pseudo option */
{ NULL, 0 } /* terminator */
};
/*
* The following is used to generate a printable argument for
* 64-bit numbers, irrespective of platform alignment and bit size.
* Because all the printf in this program use %llu as a format,
* we just return an unsigned long long, which is larger than
* we need in certain cases, but saves the hassle of using
* PRIu64 as a format specifier.
* We don't care about inlining, this is not performance critical code.
*/
unsigned long long
align_uint64(const uint64_t *pll)
{
uint64_t ret;
bcopy (pll, &ret, sizeof(ret));
return ret;
}
void *
safe_calloc(size_t number, size_t size)
{
void *ret = calloc(number, size);
if (ret == NULL)
err(EX_OSERR, "calloc");
return ret;
}
void *
safe_realloc(void *ptr, size_t size)
{
void *ret = realloc(ptr, size);
if (ret == NULL)
err(EX_OSERR, "realloc");
return ret;
}
/*
* conditionally runs the command.
* Selected options or negative -> getsockopt
*/
int
do_cmd(int optname, void *optval, uintptr_t optlen)
{
static int s = -1; /* the socket */
int i;
if (co.test_only)
return 0;
if (s == -1)
s = socket(AF_INET, SOCK_RAW, IPPROTO_RAW);
if (s < 0)
err(EX_UNAVAILABLE, "socket");
if (optname == IP_FW_GET || optname == IP_DUMMYNET_GET ||
optname == IP_FW_ADD || optname == IP_FW_TABLE_LIST ||
optname == IP_FW_TABLE_GETSIZE ||
optname == IP_FW_NAT_GET_CONFIG ||
optname < 0 ||
optname == IP_FW_NAT_GET_LOG) {
if (optname < 0)
optname = -optname;
i = getsockopt(s, IPPROTO_IP, optname, optval,
(socklen_t *)optlen);
} else {
i = setsockopt(s, IPPROTO_IP, optname, optval, optlen);
}
return i;
}
/**
* match_token takes a table and a string, returns the value associated
* with the string (-1 in case of failure).
*/
int
match_token(struct _s_x *table, char *string)
{
struct _s_x *pt;
uint i = strlen(string);
for (pt = table ; i && pt->s != NULL ; pt++)
if (strlen(pt->s) == i && !bcmp(string, pt->s, i))
return pt->x;
return -1;
}
/**
* match_value takes a table and a value, returns the string associated
* with the value (NULL in case of failure).
*/
char const *
match_value(struct _s_x *p, int value)
{
for (; p->s != NULL; p++)
if (p->x == value)
return p->s;
return NULL;
}
/*
* _substrcmp takes two strings and returns 1 if they do not match,
* and 0 if they match exactly or the first string is a sub-string
* of the second. A warning is printed to stderr in the case that the
* first string is a sub-string of the second.
*
* This function will be removed in the future through the usual
* deprecation process.
*/
int
_substrcmp(const char *str1, const char* str2)
{
if (strncmp(str1, str2, strlen(str1)) != 0)
return 1;
if (strlen(str1) != strlen(str2))
warnx("DEPRECATED: '%s' matched '%s' as a sub-string",
str1, str2);
return 0;
}
/*
* _substrcmp2 takes three strings and returns 1 if the first two do not match,
* and 0 if they match exactly or the second string is a sub-string
* of the first. A warning is printed to stderr in the case that the
* first string does not match the third.
*
* This function exists to warn about the bizzare construction
* strncmp(str, "by", 2) which is used to allow people to use a shotcut
* for "bytes". The problem is that in addition to accepting "by",
* "byt", "byte", and "bytes", it also excepts "by_rabid_dogs" and any
* other string beginning with "by".
*
* This function will be removed in the future through the usual
* deprecation process.
*/
int
_substrcmp2(const char *str1, const char* str2, const char* str3)
{
if (strncmp(str1, str2, strlen(str2)) != 0)
return 1;
if (strcmp(str1, str3) != 0)
warnx("DEPRECATED: '%s' matched '%s'",
str1, str3);
return 0;
}
/*
* prints one port, symbolic or numeric
*/
static void
print_port(int proto, uint16_t port)
{
if (proto == IPPROTO_ETHERTYPE) {
char const *s;
if (co.do_resolv && (s = match_value(ether_types, port)) )
printf("%s", s);
else
printf("0x%04x", port);
} else {
struct servent *se = NULL;
if (co.do_resolv) {
struct protoent *pe = getprotobynumber(proto);
se = getservbyport(htons(port), pe ? pe->p_name : NULL);
}
if (se)
printf("%s", se->s_name);
else
printf("%d", port);
}
}
static struct _s_x _port_name[] = {
{"dst-port", O_IP_DSTPORT},
{"src-port", O_IP_SRCPORT},
{"ipid", O_IPID},
{"iplen", O_IPLEN},
{"ipttl", O_IPTTL},
{"mac-type", O_MAC_TYPE},
{"tcpdatalen", O_TCPDATALEN},
{"tagged", O_TAGGED},
{NULL, 0}
};
/*
* Print the values in a list 16-bit items of the types above.
* XXX todo: add support for mask.
*/
static void
print_newports(ipfw_insn_u16 *cmd, int proto, int opcode)
{
uint16_t *p = cmd->ports;
int i;
char const *sep;
if (opcode != 0) {
sep = match_value(_port_name, opcode);
if (sep == NULL)
sep = "???";
printf (" %s", sep);
}
sep = " ";
for (i = F_LEN((ipfw_insn *)cmd) - 1; i > 0; i--, p += 2) {
printf("%s", sep);
print_port(proto, p[0]);
if (p[0] != p[1]) {
printf("-");
print_port(proto, p[1]);
}
sep = ",";
}
}
/*
* Like strtol, but also translates service names into port numbers
* for some protocols.
* In particular:
* proto == -1 disables the protocol check;
* proto == IPPROTO_ETHERTYPE looks up an internal table
* proto == <some value in /etc/protocols> matches the values there.
* Returns *end == s in case the parameter is not found.
*/
static int
strtoport(char *s, char **end, int base, int proto)
{
char *p, *buf;
char *s1;
int i;
*end = s; /* default - not found */
if (*s == '\0')
return 0; /* not found */
if (isdigit(*s))
return strtol(s, end, base);
/*
* find separator. '\\' escapes the next char.
*/
for (s1 = s; *s1 && (isalnum(*s1) || *s1 == '\\') ; s1++)
if (*s1 == '\\' && s1[1] != '\0')
s1++;
buf = safe_calloc(s1 - s + 1, 1);
/*
* copy into a buffer skipping backslashes
*/
for (p = s, i = 0; p != s1 ; p++)
if (*p != '\\')
buf[i++] = *p;
buf[i++] = '\0';
if (proto == IPPROTO_ETHERTYPE) {
i = match_token(ether_types, buf);
free(buf);
if (i != -1) { /* found */
*end = s1;
return i;
}
} else {
struct protoent *pe = NULL;
struct servent *se;
if (proto != 0)
pe = getprotobynumber(proto);
setservent(1);
se = getservbyname(buf, pe ? pe->p_name : NULL);
free(buf);
if (se != NULL) {
*end = s1;
return ntohs(se->s_port);
}
}
return 0; /* not found */
}
/*
* Fill the body of the command with the list of port ranges.
*/
static int
fill_newports(ipfw_insn_u16 *cmd, char *av, int proto)
{
uint16_t a, b, *p = cmd->ports;
int i = 0;
char *s = av;
while (*s) {
a = strtoport(av, &s, 0, proto);
if (s == av) /* empty or invalid argument */
return (0);
switch (*s) {
case '-': /* a range */
av = s + 1;
b = strtoport(av, &s, 0, proto);
/* Reject expressions like '1-abc' or '1-2-3'. */
if (s == av || (*s != ',' && *s != '\0'))
return (0);
p[0] = a;
p[1] = b;
break;
case ',': /* comma separated list */
case '\0':
p[0] = p[1] = a;
break;
default:
warnx("port list: invalid separator <%c> in <%s>",
*s, av);
return (0);
}
i++;
p += 2;
av = s + 1;
}
if (i > 0) {
if (i + 1 > F_LEN_MASK)
errx(EX_DATAERR, "too many ports/ranges\n");
cmd->o.len |= i + 1; /* leave F_NOT and F_OR untouched */
}
return (i);
}
static struct _s_x icmpcodes[] = {
{ "net", ICMP_UNREACH_NET },
{ "host", ICMP_UNREACH_HOST },
{ "protocol", ICMP_UNREACH_PROTOCOL },
{ "port", ICMP_UNREACH_PORT },
{ "needfrag", ICMP_UNREACH_NEEDFRAG },
{ "srcfail", ICMP_UNREACH_SRCFAIL },
{ "net-unknown", ICMP_UNREACH_NET_UNKNOWN },
{ "host-unknown", ICMP_UNREACH_HOST_UNKNOWN },
{ "isolated", ICMP_UNREACH_ISOLATED },
{ "net-prohib", ICMP_UNREACH_NET_PROHIB },
{ "host-prohib", ICMP_UNREACH_HOST_PROHIB },
{ "tosnet", ICMP_UNREACH_TOSNET },
{ "toshost", ICMP_UNREACH_TOSHOST },
{ "filter-prohib", ICMP_UNREACH_FILTER_PROHIB },
{ "host-precedence", ICMP_UNREACH_HOST_PRECEDENCE },
{ "precedence-cutoff", ICMP_UNREACH_PRECEDENCE_CUTOFF },
{ NULL, 0 }
};
static void
fill_reject_code(u_short *codep, char *str)
{
int val;
char *s;
val = strtoul(str, &s, 0);
if (s == str || *s != '\0' || val >= 0x100)
val = match_token(icmpcodes, str);
if (val < 0)
errx(EX_DATAERR, "unknown ICMP unreachable code ``%s''", str);
*codep = val;
return;
}
static void
print_reject_code(uint16_t code)
{
char const *s = match_value(icmpcodes, code);
if (s != NULL)
printf("unreach %s", s);
else
printf("unreach %u", code);
}
/*
* Returns the number of bits set (from left) in a contiguous bitmask,
* or -1 if the mask is not contiguous.
* XXX this needs a proper fix.
* This effectively works on masks in big-endian (network) format.
* when compiled on little endian architectures.
*
* First bit is bit 7 of the first byte -- note, for MAC addresses,
* the first bit on the wire is bit 0 of the first byte.
* len is the max length in bits.
*/
int
contigmask(uint8_t *p, int len)
{
int i, n;
for (i=0; i<len ; i++)
if ( (p[i/8] & (1 << (7 - (i%8)))) == 0) /* first bit unset */
break;
for (n=i+1; n < len; n++)
if ( (p[n/8] & (1 << (7 - (n%8)))) != 0)
return -1; /* mask not contiguous */
return i;
}
/*
* print flags set/clear in the two bitmasks passed as parameters.
* There is a specialized check for f_tcpflags.
*/
static void
print_flags(char const *name, ipfw_insn *cmd, struct _s_x *list)
{
char const *comma = "";
int i;
uint8_t set = cmd->arg1 & 0xff;
uint8_t clear = (cmd->arg1 >> 8) & 0xff;
if (list == f_tcpflags && set == TH_SYN && clear == TH_ACK) {
printf(" setup");
return;
}
printf(" %s ", name);
for (i=0; list[i].x != 0; i++) {
if (set & list[i].x) {
set &= ~list[i].x;
printf("%s%s", comma, list[i].s);
comma = ",";
}
if (clear & list[i].x) {
clear &= ~list[i].x;
printf("%s!%s", comma, list[i].s);
comma = ",";
}
}
}
/*
* Print the ip address contained in a command.
*/
static void
print_ip(ipfw_insn_ip *cmd, char const *s)
{
struct hostent *he = NULL;
uint32_t len = F_LEN((ipfw_insn *)cmd);
uint32_t *a = ((ipfw_insn_u32 *)cmd)->d;
if (cmd->o.opcode == O_IP_DST_LOOKUP && len > F_INSN_SIZE(ipfw_insn_u32)) {
uint32_t d = a[1];
const char *arg = "<invalid>";
if (d < sizeof(lookup_key)/sizeof(lookup_key[0]))
arg = match_value(rule_options, lookup_key[d]);
printf("%s lookup %s %d", cmd->o.len & F_NOT ? " not": "",
arg, cmd->o.arg1);
return;
}
printf("%s%s ", cmd->o.len & F_NOT ? " not": "", s);
if (cmd->o.opcode == O_IP_SRC_ME || cmd->o.opcode == O_IP_DST_ME) {
printf("me");
return;
}
if (cmd->o.opcode == O_IP_SRC_LOOKUP ||
cmd->o.opcode == O_IP_DST_LOOKUP) {
printf("table(%u", ((ipfw_insn *)cmd)->arg1);
if (len == F_INSN_SIZE(ipfw_insn_u32))
printf(",%u", *a);
printf(")");
return;
}
if (cmd->o.opcode == O_IP_SRC_SET || cmd->o.opcode == O_IP_DST_SET) {
uint32_t x, *map = (uint32_t *)&(cmd->mask);
int i, j;
char comma = '{';
x = cmd->o.arg1 - 1;
x = htonl( ~x );
cmd->addr.s_addr = htonl(cmd->addr.s_addr);
printf("%s/%d", inet_ntoa(cmd->addr),
contigmask((uint8_t *)&x, 32));
x = cmd->addr.s_addr = htonl(cmd->addr.s_addr);
x &= 0xff; /* base */
/*
* Print bits and ranges.
* Locate first bit set (i), then locate first bit unset (j).
* If we have 3+ consecutive bits set, then print them as a
* range, otherwise only print the initial bit and rescan.
*/
for (i=0; i < cmd->o.arg1; i++)
if (map[i/32] & (1<<(i & 31))) {
for (j=i+1; j < cmd->o.arg1; j++)
if (!(map[ j/32] & (1<<(j & 31))))
break;
printf("%c%d", comma, i+x);
if (j>i+2) { /* range has at least 3 elements */
printf("-%d", j-1+x);
i = j-1;
}
comma = ',';
}
printf("}");
return;
}
/*
* len == 2 indicates a single IP, whereas lists of 1 or more
* addr/mask pairs have len = (2n+1). We convert len to n so we
* use that to count the number of entries.
*/
for (len = len / 2; len > 0; len--, a += 2) {
int mb = /* mask length */
(cmd->o.opcode == O_IP_SRC || cmd->o.opcode == O_IP_DST) ?
32 : contigmask((uint8_t *)&(a[1]), 32);
if (mb == 32 && co.do_resolv)
he = gethostbyaddr((char *)&(a[0]), sizeof(u_long), AF_INET);
if (he != NULL) /* resolved to name */
printf("%s", he->h_name);
else if (mb == 0) /* any */
printf("any");
else { /* numeric IP followed by some kind of mask */
printf("%s", inet_ntoa( *((struct in_addr *)&a[0]) ) );
if (mb < 0)
printf(":%s", inet_ntoa( *((struct in_addr *)&a[1]) ) );
else if (mb < 32)
printf("/%d", mb);
}
if (len > 1)
printf(",");
}
}
/*
* prints a MAC address/mask pair
*/
static void
print_mac(uint8_t *addr, uint8_t *mask)
{
int l = contigmask(mask, 48);
if (l == 0)
printf(" any");
else {
printf(" %02x:%02x:%02x:%02x:%02x:%02x",
addr[0], addr[1], addr[2], addr[3], addr[4], addr[5]);
if (l == -1)
printf("&%02x:%02x:%02x:%02x:%02x:%02x",
mask[0], mask[1], mask[2],
mask[3], mask[4], mask[5]);
else if (l < 48)
printf("/%d", l);
}
}
static void
fill_icmptypes(ipfw_insn_u32 *cmd, char *av)
{
uint8_t type;
cmd->d[0] = 0;
while (*av) {
if (*av == ',')
av++;
type = strtoul(av, &av, 0);
if (*av != ',' && *av != '\0')
errx(EX_DATAERR, "invalid ICMP type");
if (type > 31)
errx(EX_DATAERR, "ICMP type out of range");
cmd->d[0] |= 1 << type;
}
cmd->o.opcode = O_ICMPTYPE;
cmd->o.len |= F_INSN_SIZE(ipfw_insn_u32);
}
static void
print_icmptypes(ipfw_insn_u32 *cmd)
{
int i;
char sep= ' ';
printf(" icmptypes");
for (i = 0; i < 32; i++) {
if ( (cmd->d[0] & (1 << (i))) == 0)
continue;
printf("%c%d", sep, i);
sep = ',';
}
}
/*
* show_ipfw() prints the body of an ipfw rule.
* Because the standard rule has at least proto src_ip dst_ip, we use
* a helper function to produce these entries if not provided explicitly.
* The first argument is the list of fields we have, the second is
* the list of fields we want to be printed.
*
* Special cases if we have provided a MAC header:
* + if the rule does not contain IP addresses/ports, do not print them;
* + if the rule does not contain an IP proto, print "all" instead of "ip";
*
* Once we have 'have_options', IP header fields are printed as options.
*/
#define HAVE_PROTO 0x0001
#define HAVE_SRCIP 0x0002
#define HAVE_DSTIP 0x0004
#define HAVE_PROTO4 0x0008
#define HAVE_PROTO6 0x0010
#define HAVE_OPTIONS 0x8000
#define HAVE_IP (HAVE_PROTO | HAVE_SRCIP | HAVE_DSTIP)
static void
show_prerequisites(int *flags, int want, int cmd __unused)
{
if (co.comment_only)
return;
if ( (*flags & HAVE_IP) == HAVE_IP)
*flags |= HAVE_OPTIONS;
if ( !(*flags & HAVE_OPTIONS)) {
if ( !(*flags & HAVE_PROTO) && (want & HAVE_PROTO)) {
if ( (*flags & HAVE_PROTO4))
printf(" ip4");
else if ( (*flags & HAVE_PROTO6))
printf(" ip6");
else
printf(" ip");
}
if ( !(*flags & HAVE_SRCIP) && (want & HAVE_SRCIP))
printf(" from any");
if ( !(*flags & HAVE_DSTIP) && (want & HAVE_DSTIP))
printf(" to any");
}
*flags |= want;
}
static void
show_ipfw(struct ip_fw *rule, int pcwidth, int bcwidth)
{
static int twidth = 0;
int l;
ipfw_insn *cmd, *tagptr = NULL;
const char *comment = NULL; /* ptr to comment if we have one */
int proto = 0; /* default */
int flags = 0; /* prerequisites */
ipfw_insn_log *logptr = NULL; /* set if we find an O_LOG */
ipfw_insn_altq *altqptr = NULL; /* set if we find an O_ALTQ */
int or_block = 0; /* we are in an or block */
uint32_t set_disable;
bcopy(&rule->next_rule, &set_disable, sizeof(set_disable));
if (set_disable & (1 << rule->set)) { /* disabled */
if (!co.show_sets)
return;
else
printf("# DISABLED ");
}
printf("%05u ", rule->rulenum);
if (pcwidth>0 || bcwidth>0)
printf("%*llu %*llu ", pcwidth, align_uint64(&rule->pcnt),
bcwidth, align_uint64(&rule->bcnt));
if (co.do_time == 2)
printf("%10u ", rule->timestamp);
else if (co.do_time == 1) {
char timestr[30];
time_t t = (time_t)0;
if (twidth == 0) {
strcpy(timestr, ctime(&t));
*strchr(timestr, '\n') = '\0';
twidth = strlen(timestr);
}
if (rule->timestamp) {
t = _long_to_time(rule->timestamp);
strcpy(timestr, ctime(&t));
*strchr(timestr, '\n') = '\0';
printf("%s ", timestr);
} else {
printf("%*s", twidth, " ");
}
}
if (co.show_sets)
printf("set %d ", rule->set);
/*
* print the optional "match probability"
*/
if (rule->cmd_len > 0) {
cmd = rule->cmd ;
if (cmd->opcode == O_PROB) {
ipfw_insn_u32 *p = (ipfw_insn_u32 *)cmd;
double d = 1.0 * p->d[0];
d = (d / 0x7fffffff);
printf("prob %f ", d);
}
}
/*
* first print actions
*/
for (l = rule->cmd_len - rule->act_ofs, cmd = ACTION_PTR(rule);
l > 0 ; l -= F_LEN(cmd), cmd += F_LEN(cmd)) {
switch(cmd->opcode) {
case O_CHECK_STATE:
printf("check-state");
flags = HAVE_IP; /* avoid printing anything else */
break;
case O_ACCEPT:
printf("allow");
break;
case O_COUNT:
printf("count");
break;
case O_DENY:
printf("deny");
break;
case O_REJECT:
if (cmd->arg1 == ICMP_REJECT_RST)
printf("reset");
else if (cmd->arg1 == ICMP_UNREACH_HOST)
printf("reject");
else
print_reject_code(cmd->arg1);
break;
case O_UNREACH6:
if (cmd->arg1 == ICMP6_UNREACH_RST)
printf("reset6");
else
print_unreach6_code(cmd->arg1);
break;
case O_SKIPTO:
PRINT_UINT_ARG("skipto ", cmd->arg1);
break;
case O_PIPE:
PRINT_UINT_ARG("pipe ", cmd->arg1);
break;
case O_QUEUE:
PRINT_UINT_ARG("queue ", cmd->arg1);
break;
case O_DIVERT:
PRINT_UINT_ARG("divert ", cmd->arg1);
break;
case O_TEE:
PRINT_UINT_ARG("tee ", cmd->arg1);
break;
case O_NETGRAPH:
PRINT_UINT_ARG("netgraph ", cmd->arg1);
break;
case O_NGTEE:
PRINT_UINT_ARG("ngtee ", cmd->arg1);
break;
case O_FORWARD_IP:
{
ipfw_insn_sa *s = (ipfw_insn_sa *)cmd;
if (s->sa.sin_addr.s_addr == INADDR_ANY) {
printf("fwd tablearg");
} else {
printf("fwd %s", inet_ntoa(s->sa.sin_addr));
}
if (s->sa.sin_port)
printf(",%d", s->sa.sin_port);
}
break;
case O_LOG: /* O_LOG is printed last */
logptr = (ipfw_insn_log *)cmd;
break;
case O_ALTQ: /* O_ALTQ is printed after O_LOG */
altqptr = (ipfw_insn_altq *)cmd;
break;
case O_TAG:
tagptr = cmd;
break;
case O_NAT:
PRINT_UINT_ARG("nat ", cmd->arg1);
break;
case O_SETFIB:
PRINT_UINT_ARG("setfib ", cmd->arg1);
break;
case O_REASS:
printf("reass");
break;
default:
printf("** unrecognized action %d len %d ",
cmd->opcode, cmd->len);
}
}
if (logptr) {
if (logptr->max_log > 0)
printf(" log logamount %d", logptr->max_log);
else
printf(" log");
}
#ifndef NO_ALTQ
if (altqptr) {
print_altq_cmd(altqptr);
}
#endif
if (tagptr) {
if (tagptr->len & F_NOT)
PRINT_UINT_ARG(" untag ", tagptr->arg1);
else
PRINT_UINT_ARG(" tag ", tagptr->arg1);
}
/*
* then print the body.
*/
for (l = rule->act_ofs, cmd = rule->cmd ;
l > 0 ; l -= F_LEN(cmd) , cmd += F_LEN(cmd)) {
if ((cmd->len & F_OR) || (cmd->len & F_NOT))
continue;
if (cmd->opcode == O_IP4) {
flags |= HAVE_PROTO4;
break;
} else if (cmd->opcode == O_IP6) {
flags |= HAVE_PROTO6;
break;
}
}
if (rule->_pad & 1) { /* empty rules before options */
if (!co.do_compact) {
show_prerequisites(&flags, HAVE_PROTO, 0);
printf(" from any to any");
}
flags |= HAVE_IP | HAVE_OPTIONS;
}
if (co.comment_only)
comment = "...";
for (l = rule->act_ofs, cmd = rule->cmd ;
l > 0 ; l -= F_LEN(cmd) , cmd += F_LEN(cmd)) {
/* useful alias */
ipfw_insn_u32 *cmd32 = (ipfw_insn_u32 *)cmd;
if (co.comment_only) {
if (cmd->opcode != O_NOP)
continue;
printf(" // %s\n", (char *)(cmd + 1));
return;
}
show_prerequisites(&flags, 0, cmd->opcode);
switch(cmd->opcode) {
case O_PROB:
break; /* done already */
case O_PROBE_STATE:
break; /* no need to print anything here */
case O_IP_SRC:
case O_IP_SRC_LOOKUP:
case O_IP_SRC_MASK:
case O_IP_SRC_ME:
case O_IP_SRC_SET:
show_prerequisites(&flags, HAVE_PROTO, 0);
if (!(flags & HAVE_SRCIP))
printf(" from");
if ((cmd->len & F_OR) && !or_block)
printf(" {");
print_ip((ipfw_insn_ip *)cmd,
(flags & HAVE_OPTIONS) ? " src-ip" : "");
flags |= HAVE_SRCIP;
break;
case O_IP_DST:
case O_IP_DST_LOOKUP:
case O_IP_DST_MASK:
case O_IP_DST_ME:
case O_IP_DST_SET:
show_prerequisites(&flags, HAVE_PROTO|HAVE_SRCIP, 0);
if (!(flags & HAVE_DSTIP))
printf(" to");
if ((cmd->len & F_OR) && !or_block)
printf(" {");
print_ip((ipfw_insn_ip *)cmd,
(flags & HAVE_OPTIONS) ? " dst-ip" : "");
flags |= HAVE_DSTIP;
break;
case O_IP6_SRC:
case O_IP6_SRC_MASK:
case O_IP6_SRC_ME:
show_prerequisites(&flags, HAVE_PROTO, 0);
if (!(flags & HAVE_SRCIP))
printf(" from");
if ((cmd->len & F_OR) && !or_block)
printf(" {");
print_ip6((ipfw_insn_ip6 *)cmd,
(flags & HAVE_OPTIONS) ? " src-ip6" : "");
flags |= HAVE_SRCIP | HAVE_PROTO;
break;
case O_IP6_DST:
case O_IP6_DST_MASK:
case O_IP6_DST_ME:
show_prerequisites(&flags, HAVE_PROTO|HAVE_SRCIP, 0);
if (!(flags & HAVE_DSTIP))
printf(" to");
if ((cmd->len & F_OR) && !or_block)
printf(" {");
print_ip6((ipfw_insn_ip6 *)cmd,
(flags & HAVE_OPTIONS) ? " dst-ip6" : "");
flags |= HAVE_DSTIP;
break;
case O_FLOW6ID:
print_flow6id( (ipfw_insn_u32 *) cmd );
flags |= HAVE_OPTIONS;
break;
case O_IP_DSTPORT:
show_prerequisites(&flags, HAVE_IP, 0);
case O_IP_SRCPORT:
show_prerequisites(&flags, HAVE_PROTO|HAVE_SRCIP, 0);
if ((cmd->len & F_OR) && !or_block)
printf(" {");
if (cmd->len & F_NOT)
printf(" not");
print_newports((ipfw_insn_u16 *)cmd, proto,
(flags & HAVE_OPTIONS) ? cmd->opcode : 0);
break;
case O_PROTO: {
struct protoent *pe = NULL;
if ((cmd->len & F_OR) && !or_block)
printf(" {");
if (cmd->len & F_NOT)
printf(" not");
proto = cmd->arg1;
pe = getprotobynumber(cmd->arg1);
if ((flags & (HAVE_PROTO4 | HAVE_PROTO6)) &&
!(flags & HAVE_PROTO))
show_prerequisites(&flags,
HAVE_IP | HAVE_OPTIONS, 0);
if (flags & HAVE_OPTIONS)
printf(" proto");
if (pe)
printf(" %s", pe->p_name);
else
printf(" %u", cmd->arg1);
}
flags |= HAVE_PROTO;
break;
default: /*options ... */
if (!(cmd->len & (F_OR|F_NOT)))
if (((cmd->opcode == O_IP6) &&
(flags & HAVE_PROTO6)) ||
((cmd->opcode == O_IP4) &&
(flags & HAVE_PROTO4)))
break;
show_prerequisites(&flags, HAVE_IP | HAVE_OPTIONS, 0);
if ((cmd->len & F_OR) && !or_block)
printf(" {");
if (cmd->len & F_NOT && cmd->opcode != O_IN)
printf(" not");
switch(cmd->opcode) {
case O_MACADDR2: {
ipfw_insn_mac *m = (ipfw_insn_mac *)cmd;
printf(" MAC");
print_mac(m->addr, m->mask);
print_mac(m->addr + 6, m->mask + 6);
}
break;
case O_MAC_TYPE:
print_newports((ipfw_insn_u16 *)cmd,
IPPROTO_ETHERTYPE, cmd->opcode);
break;
case O_FRAG:
printf(" frag");
break;
case O_FIB:
printf(" fib %u", cmd->arg1 );
break;
case O_IN:
printf(cmd->len & F_NOT ? " out" : " in");
break;
case O_DIVERTED:
switch (cmd->arg1) {
case 3:
printf(" diverted");
break;
case 1:
printf(" diverted-loopback");
break;
case 2:
printf(" diverted-output");
break;
default:
printf(" diverted-?<%u>", cmd->arg1);
break;
}
break;
case O_LAYER2:
printf(" layer2");
break;
case O_XMIT:
case O_RECV:
case O_VIA:
{
char const *s;
ipfw_insn_if *cmdif = (ipfw_insn_if *)cmd;
if (cmd->opcode == O_XMIT)
s = "xmit";
else if (cmd->opcode == O_RECV)
s = "recv";
else /* if (cmd->opcode == O_VIA) */
s = "via";
if (cmdif->name[0] == '\0')
printf(" %s %s", s,
inet_ntoa(cmdif->p.ip));
else
printf(" %s %s", s, cmdif->name);
break;
}
case O_IPID:
if (F_LEN(cmd) == 1)
printf(" ipid %u", cmd->arg1 );
else
print_newports((ipfw_insn_u16 *)cmd, 0,
O_IPID);
break;
case O_IPTTL:
if (F_LEN(cmd) == 1)
printf(" ipttl %u", cmd->arg1 );
else
print_newports((ipfw_insn_u16 *)cmd, 0,
O_IPTTL);
break;
case O_IPVER:
printf(" ipver %u", cmd->arg1 );
break;
case O_IPPRECEDENCE:
printf(" ipprecedence %u", (cmd->arg1) >> 5 );
break;
case O_IPLEN:
if (F_LEN(cmd) == 1)
printf(" iplen %u", cmd->arg1 );
else
print_newports((ipfw_insn_u16 *)cmd, 0,
O_IPLEN);
break;
case O_IPOPT:
print_flags("ipoptions", cmd, f_ipopts);
break;
case O_IPTOS:
print_flags("iptos", cmd, f_iptos);
break;
case O_ICMPTYPE:
print_icmptypes((ipfw_insn_u32 *)cmd);
break;
case O_ESTAB:
printf(" established");
break;
case O_TCPDATALEN:
if (F_LEN(cmd) == 1)
printf(" tcpdatalen %u", cmd->arg1 );
else
print_newports((ipfw_insn_u16 *)cmd, 0,
O_TCPDATALEN);
break;
case O_TCPFLAGS:
print_flags("tcpflags", cmd, f_tcpflags);
break;
case O_TCPOPTS:
print_flags("tcpoptions", cmd, f_tcpopts);
break;
case O_TCPWIN:
printf(" tcpwin %d", ntohs(cmd->arg1));
break;
case O_TCPACK:
printf(" tcpack %d", ntohl(cmd32->d[0]));
break;
case O_TCPSEQ:
printf(" tcpseq %d", ntohl(cmd32->d[0]));
break;
case O_UID:
{
struct passwd *pwd = getpwuid(cmd32->d[0]);
if (pwd)
printf(" uid %s", pwd->pw_name);
else
printf(" uid %u", cmd32->d[0]);
}
break;
case O_GID:
{
struct group *grp = getgrgid(cmd32->d[0]);
if (grp)
printf(" gid %s", grp->gr_name);
else
printf(" gid %u", cmd32->d[0]);
}
break;
case O_JAIL:
printf(" jail %d", cmd32->d[0]);
break;
case O_VERREVPATH:
printf(" verrevpath");
break;
case O_VERSRCREACH:
printf(" versrcreach");
break;
case O_ANTISPOOF:
printf(" antispoof");
break;
case O_IPSEC:
printf(" ipsec");
break;
case O_NOP:
comment = (char *)(cmd + 1);
break;
case O_KEEP_STATE:
printf(" keep-state");
break;
case O_LIMIT: {
struct _s_x *p = limit_masks;
ipfw_insn_limit *c = (ipfw_insn_limit *)cmd;
uint8_t x = c->limit_mask;
char const *comma = " ";
printf(" limit");
for (; p->x != 0 ; p++)
if ((x & p->x) == p->x) {
x &= ~p->x;
printf("%s%s", comma, p->s);
comma = ",";
}
PRINT_UINT_ARG(" ", c->conn_limit);
break;
}
case O_IP6:
printf(" ip6");
break;
case O_IP4:
printf(" ip4");
break;
case O_ICMP6TYPE:
print_icmp6types((ipfw_insn_u32 *)cmd);
break;
case O_EXT_HDR:
print_ext6hdr( (ipfw_insn *) cmd );
break;
case O_TAGGED:
if (F_LEN(cmd) == 1)
PRINT_UINT_ARG(" tagged ", cmd->arg1);
else
print_newports((ipfw_insn_u16 *)cmd, 0,
O_TAGGED);
break;
default:
printf(" [opcode %d len %d]",
cmd->opcode, cmd->len);
}
}
if (cmd->len & F_OR) {
printf(" or");
or_block = 1;
} else if (or_block) {
printf(" }");
or_block = 0;
}
}
show_prerequisites(&flags, HAVE_IP, 0);
if (comment)
printf(" // %s", comment);
printf("\n");
}
static void
show_dyn_ipfw(ipfw_dyn_rule *d, int pcwidth, int bcwidth)
{
struct protoent *pe;
struct in_addr a;
uint16_t rulenum;
char buf[INET6_ADDRSTRLEN];
if (!co.do_expired) {
if (!d->expire && !(d->dyn_type == O_LIMIT_PARENT))
return;
}
bcopy(&d->rule, &rulenum, sizeof(rulenum));
printf("%05d", rulenum);
if (pcwidth>0 || bcwidth>0)
printf(" %*llu %*llu (%ds)", pcwidth,
align_uint64(&d->pcnt), bcwidth,
align_uint64(&d->bcnt), d->expire);
switch (d->dyn_type) {
case O_LIMIT_PARENT:
printf(" PARENT %d", d->count);
break;
case O_LIMIT:
printf(" LIMIT");
break;
case O_KEEP_STATE: /* bidir, no mask */
printf(" STATE");
break;
}
if ((pe = getprotobynumber(d->id.proto)) != NULL)
printf(" %s", pe->p_name);
else
printf(" proto %u", d->id.proto);
if (d->id.addr_type == 4) {
a.s_addr = htonl(d->id.src_ip);
printf(" %s %d", inet_ntoa(a), d->id.src_port);
a.s_addr = htonl(d->id.dst_ip);
printf(" <-> %s %d", inet_ntoa(a), d->id.dst_port);
} else if (d->id.addr_type == 6) {
printf(" %s %d", inet_ntop(AF_INET6, &d->id.src_ip6, buf,
sizeof(buf)), d->id.src_port);
printf(" <-> %s %d", inet_ntop(AF_INET6, &d->id.dst_ip6, buf,
sizeof(buf)), d->id.dst_port);
} else
printf(" UNKNOWN <-> UNKNOWN\n");
printf("\n");
}
/*
* This one handles all set-related commands
* ipfw set { show | enable | disable }
* ipfw set swap X Y
* ipfw set move X to Y
* ipfw set move rule X to Y
*/
void
ipfw_sets_handler(char *av[])
{
uint32_t set_disable, masks[2];
int i, nbytes;
uint16_t rulenum;
uint8_t cmd, new_set;
av++;
if (av[0] == NULL)
errx(EX_USAGE, "set needs command");
if (_substrcmp(*av, "show") == 0) {
void *data = NULL;
char const *msg;
int nalloc;
nalloc = nbytes = sizeof(struct ip_fw);
while (nbytes >= nalloc) {
if (data)
free(data);
nalloc = nalloc * 2 + 200;
nbytes = nalloc;
data = safe_calloc(1, nbytes);
if (do_cmd(IP_FW_GET, data, (uintptr_t)&nbytes) < 0)
err(EX_OSERR, "getsockopt(IP_FW_GET)");
}
bcopy(&((struct ip_fw *)data)->next_rule,
&set_disable, sizeof(set_disable));
for (i = 0, msg = "disable" ; i < RESVD_SET; i++)
if ((set_disable & (1<<i))) {
printf("%s %d", msg, i);
msg = "";
}
msg = (set_disable) ? " enable" : "enable";
for (i = 0; i < RESVD_SET; i++)
if (!(set_disable & (1<<i))) {
printf("%s %d", msg, i);
msg = "";
}
printf("\n");
} else if (_substrcmp(*av, "swap") == 0) {
av++;
if ( av[0] == NULL || av[1] == NULL )
errx(EX_USAGE, "set swap needs 2 set numbers\n");
rulenum = atoi(av[0]);
new_set = atoi(av[1]);
if (!isdigit(*(av[0])) || rulenum > RESVD_SET)
errx(EX_DATAERR, "invalid set number %s\n", av[0]);
if (!isdigit(*(av[1])) || new_set > RESVD_SET)
errx(EX_DATAERR, "invalid set number %s\n", av[1]);
masks[0] = (4 << 24) | (new_set << 16) | (rulenum);
i = do_cmd(IP_FW_DEL, masks, sizeof(uint32_t));
} else if (_substrcmp(*av, "move") == 0) {
av++;
if (av[0] && _substrcmp(*av, "rule") == 0) {
cmd = 2;
av++;
} else
cmd = 3;
if (av[0] == NULL || av[1] == NULL || av[2] == NULL ||
av[3] != NULL || _substrcmp(av[1], "to") != 0)
errx(EX_USAGE, "syntax: set move [rule] X to Y\n");
rulenum = atoi(av[0]);
new_set = atoi(av[2]);
if (!isdigit(*(av[0])) || (cmd == 3 && rulenum > RESVD_SET) ||
(cmd == 2 && rulenum == IPFW_DEFAULT_RULE) )
errx(EX_DATAERR, "invalid source number %s\n", av[0]);
if (!isdigit(*(av[2])) || new_set > RESVD_SET)
errx(EX_DATAERR, "invalid dest. set %s\n", av[1]);
masks[0] = (cmd << 24) | (new_set << 16) | (rulenum);
i = do_cmd(IP_FW_DEL, masks, sizeof(uint32_t));
} else if (_substrcmp(*av, "disable") == 0 ||
_substrcmp(*av, "enable") == 0 ) {
int which = _substrcmp(*av, "enable") == 0 ? 1 : 0;
av++;
masks[0] = masks[1] = 0;
while (av[0]) {
if (isdigit(**av)) {
i = atoi(*av);
if (i < 0 || i > RESVD_SET)
errx(EX_DATAERR,
"invalid set number %d\n", i);
masks[which] |= (1<<i);
} else if (_substrcmp(*av, "disable") == 0)
which = 0;
else if (_substrcmp(*av, "enable") == 0)
which = 1;
else
errx(EX_DATAERR,
"invalid set command %s\n", *av);
av++;
}
if ( (masks[0] & masks[1]) != 0 )
errx(EX_DATAERR,
"cannot enable and disable the same set\n");
i = do_cmd(IP_FW_DEL, masks, sizeof(masks));
if (i)
warn("set enable/disable: setsockopt(IP_FW_DEL)");
} else
errx(EX_USAGE, "invalid set command %s\n", *av);
}
void
ipfw_sysctl_handler(char *av[], int which)
{
av++;
if (av[0] == NULL) {
warnx("missing keyword to enable/disable\n");
} else if (_substrcmp(*av, "firewall") == 0) {
sysctlbyname("net.inet.ip.fw.enable", NULL, 0,
&which, sizeof(which));
} else if (_substrcmp(*av, "one_pass") == 0) {
sysctlbyname("net.inet.ip.fw.one_pass", NULL, 0,
&which, sizeof(which));
} else if (_substrcmp(*av, "debug") == 0) {
sysctlbyname("net.inet.ip.fw.debug", NULL, 0,
&which, sizeof(which));
} else if (_substrcmp(*av, "verbose") == 0) {
sysctlbyname("net.inet.ip.fw.verbose", NULL, 0,
&which, sizeof(which));
} else if (_substrcmp(*av, "dyn_keepalive") == 0) {
sysctlbyname("net.inet.ip.fw.dyn_keepalive", NULL, 0,
&which, sizeof(which));
#ifndef NO_ALTQ
} else if (_substrcmp(*av, "altq") == 0) {
altq_set_enabled(which);
#endif
} else {
warnx("unrecognize enable/disable keyword: %s\n", *av);
}
}
void
ipfw_list(int ac, char *av[], int show_counters)
{
struct ip_fw *r;
ipfw_dyn_rule *dynrules, *d;
#define NEXT(r) ((struct ip_fw *)((char *)r + RULESIZE(r)))
char *lim;
void *data = NULL;
int bcwidth, n, nbytes, nstat, ndyn, pcwidth, width;
int exitval = EX_OK;
int lac;
char **lav;
u_long rnum, last;
char *endptr;
int seen = 0;
uint8_t set;
const int ocmd = co.do_pipe ? IP_DUMMYNET_GET : IP_FW_GET;
int nalloc = 1024; /* start somewhere... */
last = 0;
if (co.test_only) {
fprintf(stderr, "Testing only, list disabled\n");
return;
}
if (co.do_pipe) {
dummynet_list(ac, av, show_counters);
return;
}
ac--;
av++;
/* get rules or pipes from kernel, resizing array as necessary */
nbytes = nalloc;
while (nbytes >= nalloc) {
nalloc = nalloc * 2 + 200;
nbytes = nalloc;
data = safe_realloc(data, nbytes);
if (do_cmd(ocmd, data, (uintptr_t)&nbytes) < 0)
err(EX_OSERR, "getsockopt(IP_%s_GET)",
co.do_pipe ? "DUMMYNET" : "FW");
}
/*
* Count static rules. They have variable size so we
* need to scan the list to count them.
*/
for (nstat = 1, r = data, lim = (char *)data + nbytes;
r->rulenum < IPFW_DEFAULT_RULE && (char *)r < lim;
++nstat, r = NEXT(r) )
; /* nothing */
/*
* Count dynamic rules. This is easier as they have
* fixed size.
*/
r = NEXT(r);
dynrules = (ipfw_dyn_rule *)r ;
n = (char *)r - (char *)data;
ndyn = (nbytes - n) / sizeof *dynrules;
/* if showing stats, figure out column widths ahead of time */
bcwidth = pcwidth = 0;
if (show_counters) {
for (n = 0, r = data; n < nstat; n++, r = NEXT(r)) {
/* skip rules from another set */
if (co.use_set && r->set != co.use_set - 1)
continue;
/* packet counter */
width = snprintf(NULL, 0, "%llu",
align_uint64(&r->pcnt));
if (width > pcwidth)
pcwidth = width;
/* byte counter */
width = snprintf(NULL, 0, "%llu",
align_uint64(&r->bcnt));
if (width > bcwidth)
bcwidth = width;
}
}
if (co.do_dynamic && ndyn) {
for (n = 0, d = dynrules; n < ndyn; n++, d++) {
if (co.use_set) {
/* skip rules from another set */
bcopy((char *)&d->rule + sizeof(uint16_t),
&set, sizeof(uint8_t));
if (set != co.use_set - 1)
continue;
}
width = snprintf(NULL, 0, "%llu",
align_uint64(&d->pcnt));
if (width > pcwidth)
pcwidth = width;
width = snprintf(NULL, 0, "%llu",
align_uint64(&d->bcnt));
if (width > bcwidth)
bcwidth = width;
}
}
/* if no rule numbers were specified, list all rules */
if (ac == 0) {
for (n = 0, r = data; n < nstat; n++, r = NEXT(r)) {
if (co.use_set && r->set != co.use_set - 1)
continue;
show_ipfw(r, pcwidth, bcwidth);
}
if (co.do_dynamic && ndyn) {
printf("## Dynamic rules (%d):\n", ndyn);
for (n = 0, d = dynrules; n < ndyn; n++, d++) {
if (co.use_set) {
bcopy((char *)&d->rule + sizeof(uint16_t),
&set, sizeof(uint8_t));
if (set != co.use_set - 1)
continue;
}
show_dyn_ipfw(d, pcwidth, bcwidth);
}
}
goto done;
}
/* display specific rules requested on command line */
for (lac = ac, lav = av; lac != 0; lac--) {
/* convert command line rule # */
last = rnum = strtoul(*lav++, &endptr, 10);
if (*endptr == '-')
last = strtoul(endptr+1, &endptr, 10);
if (*endptr) {
exitval = EX_USAGE;
warnx("invalid rule number: %s", *(lav - 1));
continue;
}
for (n = seen = 0, r = data; n < nstat; n++, r = NEXT(r) ) {
if (r->rulenum > last)
break;
if (co.use_set && r->set != co.use_set - 1)
continue;
if (r->rulenum >= rnum && r->rulenum <= last) {
show_ipfw(r, pcwidth, bcwidth);
seen = 1;
}
}
if (!seen) {
/* give precedence to other error(s) */
if (exitval == EX_OK)
exitval = EX_UNAVAILABLE;
warnx("rule %lu does not exist", rnum);
}
}
if (co.do_dynamic && ndyn) {
printf("## Dynamic rules:\n");
for (lac = ac, lav = av; lac != 0; lac--) {
last = rnum = strtoul(*lav++, &endptr, 10);
if (*endptr == '-')
last = strtoul(endptr+1, &endptr, 10);
if (*endptr)
/* already warned */
continue;
for (n = 0, d = dynrules; n < ndyn; n++, d++) {
uint16_t rulenum;
bcopy(&d->rule, &rulenum, sizeof(rulenum));
if (rulenum > rnum)
break;
if (co.use_set) {
bcopy((char *)&d->rule + sizeof(uint16_t),
&set, sizeof(uint8_t));
if (set != co.use_set - 1)
continue;
}
if (r->rulenum >= rnum && r->rulenum <= last)
show_dyn_ipfw(d, pcwidth, bcwidth);
}
}
}
ac = 0;
done:
free(data);
if (exitval != EX_OK)
exit(exitval);
#undef NEXT
}
static int
lookup_host (char *host, struct in_addr *ipaddr)
{
struct hostent *he;
if (!inet_aton(host, ipaddr)) {
if ((he = gethostbyname(host)) == NULL)
return(-1);
*ipaddr = *(struct in_addr *)he->h_addr_list[0];
}
return(0);
}
/*
* fills the addr and mask fields in the instruction as appropriate from av.
* Update length as appropriate.
* The following formats are allowed:
* me returns O_IP_*_ME
* 1.2.3.4 single IP address
* 1.2.3.4:5.6.7.8 address:mask
* 1.2.3.4/24 address/mask
* 1.2.3.4/26{1,6,5,4,23} set of addresses in a subnet
* We can have multiple comma-separated address/mask entries.
*/
static void
fill_ip(ipfw_insn_ip *cmd, char *av)
{
int len = 0;
uint32_t *d = ((ipfw_insn_u32 *)cmd)->d;
cmd->o.len &= ~F_LEN_MASK; /* zero len */
if (_substrcmp(av, "any") == 0)
return;
if (_substrcmp(av, "me") == 0) {
cmd->o.len |= F_INSN_SIZE(ipfw_insn);
return;
}
if (strncmp(av, "table(", 6) == 0) {
char *p = strchr(av + 6, ',');
if (p)
*p++ = '\0';
cmd->o.opcode = O_IP_DST_LOOKUP;
cmd->o.arg1 = strtoul(av + 6, NULL, 0);
if (p) {
cmd->o.len |= F_INSN_SIZE(ipfw_insn_u32);
d[0] = strtoul(p, NULL, 0);
} else
cmd->o.len |= F_INSN_SIZE(ipfw_insn);
return;
}
while (av) {
/*
* After the address we can have '/' or ':' indicating a mask,
* ',' indicating another address follows, '{' indicating a
* set of addresses of unspecified size.
*/
char *t = NULL, *p = strpbrk(av, "/:,{");
int masklen;
char md, nd = '\0';
if (p) {
md = *p;
*p++ = '\0';
if ((t = strpbrk(p, ",{")) != NULL) {
nd = *t;
*t = '\0';
}
} else
md = '\0';
if (lookup_host(av, (struct in_addr *)&d[0]) != 0)
errx(EX_NOHOST, "hostname ``%s'' unknown", av);
switch (md) {
case ':':
if (!inet_aton(p, (struct in_addr *)&d[1]))
errx(EX_DATAERR, "bad netmask ``%s''", p);
break;
case '/':
masklen = atoi(p);
if (masklen == 0)
d[1] = htonl(0); /* mask */
else if (masklen > 32)
errx(EX_DATAERR, "bad width ``%s''", p);
else
d[1] = htonl(~0 << (32 - masklen));
break;
case '{': /* no mask, assume /24 and put back the '{' */
d[1] = htonl(~0 << (32 - 24));
*(--p) = md;
break;
case ',': /* single address plus continuation */
*(--p) = md;
/* FALLTHROUGH */
case 0: /* initialization value */
default:
d[1] = htonl(~0); /* force /32 */
break;
}
d[0] &= d[1]; /* mask base address with mask */
if (t)
*t = nd;
/* find next separator */
if (p)
p = strpbrk(p, ",{");
if (p && *p == '{') {
/*
* We have a set of addresses. They are stored as follows:
* arg1 is the set size (powers of 2, 2..256)
* addr is the base address IN HOST FORMAT
* mask.. is an array of arg1 bits (rounded up to
* the next multiple of 32) with bits set
* for each host in the map.
*/
uint32_t *map = (uint32_t *)&cmd->mask;
int low, high;
int i = contigmask((uint8_t *)&(d[1]), 32);
if (len > 0)
errx(EX_DATAERR, "address set cannot be in a list");
if (i < 24 || i > 31)
errx(EX_DATAERR, "invalid set with mask %d\n", i);
cmd->o.arg1 = 1<<(32-i); /* map length */
d[0] = ntohl(d[0]); /* base addr in host format */
cmd->o.opcode = O_IP_DST_SET; /* default */
cmd->o.len |= F_INSN_SIZE(ipfw_insn_u32) + (cmd->o.arg1+31)/32;
for (i = 0; i < (cmd->o.arg1+31)/32 ; i++)
map[i] = 0; /* clear map */
av = p + 1;
low = d[0] & 0xff;
high = low + cmd->o.arg1 - 1;
/*
* Here, i stores the previous value when we specify a range
* of addresses within a mask, e.g. 45-63. i = -1 means we
* have no previous value.
*/
i = -1; /* previous value in a range */
while (isdigit(*av)) {
char *s;
int a = strtol(av, &s, 0);
if (s == av) { /* no parameter */
if (*av != '}')
errx(EX_DATAERR, "set not closed\n");
if (i != -1)
errx(EX_DATAERR, "incomplete range %d-", i);
break;
}
if (a < low || a > high)
errx(EX_DATAERR, "addr %d out of range [%d-%d]\n",
a, low, high);
a -= low;
if (i == -1) /* no previous in range */
i = a;
else { /* check that range is valid */
if (i > a)
errx(EX_DATAERR, "invalid range %d-%d",
i+low, a+low);
if (*s == '-')
errx(EX_DATAERR, "double '-' in range");
}
for (; i <= a; i++)
map[i/32] |= 1<<(i & 31);
i = -1;
if (*s == '-')
i = a;
else if (*s == '}')
break;
av = s+1;
}
return;
}
av = p;
if (av) /* then *av must be a ',' */
av++;
/* Check this entry */
if (d[1] == 0) { /* "any", specified as x.x.x.x/0 */
/*
* 'any' turns the entire list into a NOP.
* 'not any' never matches, so it is removed from the
* list unless it is the only item, in which case we
* report an error.
*/
if (cmd->o.len & F_NOT) { /* "not any" never matches */
if (av == NULL && len == 0) /* only this entry */
errx(EX_DATAERR, "not any never matches");
}
/* else do nothing and skip this entry */
return;
}
/* A single IP can be stored in an optimized format */
if (d[1] == (uint32_t)~0 && av == NULL && len == 0) {
cmd->o.len |= F_INSN_SIZE(ipfw_insn_u32);
return;
}
len += 2; /* two words... */
d += 2;
} /* end while */
if (len + 1 > F_LEN_MASK)
errx(EX_DATAERR, "address list too long");
cmd->o.len |= len+1;
}
/* n2mask sets n bits of the mask */
void
n2mask(struct in6_addr *mask, int n)
{
static int minimask[9] =
{ 0x00, 0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff };
u_char *p;
memset(mask, 0, sizeof(struct in6_addr));
p = (u_char *) mask;
for (; n > 0; p++, n -= 8) {
if (n >= 8)
*p = 0xff;
else
*p = minimask[n];
}
return;
}
/*
* helper function to process a set of flags and set bits in the
* appropriate masks.
*/
static void
fill_flags(ipfw_insn *cmd, enum ipfw_opcodes opcode,
struct _s_x *flags, char *p)
{
uint8_t set=0, clear=0;
while (p && *p) {
char *q; /* points to the separator */
int val;
uint8_t *which; /* mask we are working on */
if (*p == '!') {
p++;
which = &clear;
} else
which = &set;
q = strchr(p, ',');
if (q)
*q++ = '\0';
val = match_token(flags, p);
if (val <= 0)
errx(EX_DATAERR, "invalid flag %s", p);
*which |= (uint8_t)val;
p = q;
}
cmd->opcode = opcode;
cmd->len = (cmd->len & (F_NOT | F_OR)) | 1;
cmd->arg1 = (set & 0xff) | ( (clear & 0xff) << 8);
}
void
ipfw_delete(char *av[])
{
uint32_t rulenum;
int i;
int exitval = EX_OK;
int do_set = 0;
av++;
NEED1("missing rule specification");
if ( *av && _substrcmp(*av, "set") == 0) {
/* Do not allow using the following syntax:
* ipfw set N delete set M
*/
if (co.use_set)
errx(EX_DATAERR, "invalid syntax");
do_set = 1; /* delete set */
av++;
}
/* Rule number */
while (*av && isdigit(**av)) {
i = atoi(*av); av++;
if (co.do_nat) {
exitval = do_cmd(IP_FW_NAT_DEL, &i, sizeof i);
if (exitval) {
exitval = EX_UNAVAILABLE;
warn("rule %u not available", i);
}
} else if (co.do_pipe) {
exitval = ipfw_delete_pipe(co.do_pipe, i);
} else {
if (co.use_set)
rulenum = (i & 0xffff) | (5 << 24) |
((co.use_set - 1) << 16);
else
rulenum = (i & 0xffff) | (do_set << 24);
i = do_cmd(IP_FW_DEL, &rulenum, sizeof rulenum);
if (i) {
exitval = EX_UNAVAILABLE;
warn("rule %u: setsockopt(IP_FW_DEL)",
rulenum);
}
}
}
if (exitval != EX_OK)
exit(exitval);
}
/*
* fill the interface structure. We do not check the name as we can
* create interfaces dynamically, so checking them at insert time
* makes relatively little sense.
* Interface names containing '*', '?', or '[' are assumed to be shell
* patterns which match interfaces.
*/
static void
fill_iface(ipfw_insn_if *cmd, char *arg)
{
cmd->name[0] = '\0';
cmd->o.len |= F_INSN_SIZE(ipfw_insn_if);
/* Parse the interface or address */
if (strcmp(arg, "any") == 0)
cmd->o.len = 0; /* effectively ignore this command */
else if (!isdigit(*arg)) {
strlcpy(cmd->name, arg, sizeof(cmd->name));
cmd->p.glob = strpbrk(arg, "*?[") != NULL ? 1 : 0;
} else if (!inet_aton(arg, &cmd->p.ip))
errx(EX_DATAERR, "bad ip address ``%s''", arg);
}
static void
get_mac_addr_mask(const char *p, uint8_t *addr, uint8_t *mask)
{
int i;
size_t l;
char *ap, *ptr, *optr;
struct ether_addr *mac;
const char *macset = "0123456789abcdefABCDEF:";
if (strcmp(p, "any") == 0) {
for (i = 0; i < ETHER_ADDR_LEN; i++)
addr[i] = mask[i] = 0;
return;
}
optr = ptr = strdup(p);
if ((ap = strsep(&ptr, "&/")) != NULL && *ap != 0) {
l = strlen(ap);
if (strspn(ap, macset) != l || (mac = ether_aton(ap)) == NULL)
errx(EX_DATAERR, "Incorrect MAC address");
bcopy(mac, addr, ETHER_ADDR_LEN);
} else
errx(EX_DATAERR, "Incorrect MAC address");
if (ptr != NULL) { /* we have mask? */
if (p[ptr - optr - 1] == '/') { /* mask len */
long ml = strtol(ptr, &ap, 10);
if (*ap != 0 || ml > ETHER_ADDR_LEN * 8 || ml < 0)
errx(EX_DATAERR, "Incorrect mask length");
for (i = 0; ml > 0 && i < ETHER_ADDR_LEN; ml -= 8, i++)
mask[i] = (ml >= 8) ? 0xff: (~0) << (8 - ml);
} else { /* mask */
l = strlen(ptr);
if (strspn(ptr, macset) != l ||
(mac = ether_aton(ptr)) == NULL)
errx(EX_DATAERR, "Incorrect mask");
bcopy(mac, mask, ETHER_ADDR_LEN);
}
} else { /* default mask: ff:ff:ff:ff:ff:ff */
for (i = 0; i < ETHER_ADDR_LEN; i++)
mask[i] = 0xff;
}
for (i = 0; i < ETHER_ADDR_LEN; i++)
addr[i] &= mask[i];
free(optr);
}
/*
* helper function, updates the pointer to cmd with the length
* of the current command, and also cleans up the first word of
* the new command in case it has been clobbered before.
*/
static ipfw_insn *
next_cmd(ipfw_insn *cmd)
{
cmd += F_LEN(cmd);
bzero(cmd, sizeof(*cmd));
return cmd;
}
/*
* Takes arguments and copies them into a comment
*/
static void
fill_comment(ipfw_insn *cmd, char **av)
{
int i, l;
char *p = (char *)(cmd + 1);
cmd->opcode = O_NOP;
cmd->len = (cmd->len & (F_NOT | F_OR));
/* Compute length of comment string. */
for (i = 0, l = 0; av[i] != NULL; i++)
l += strlen(av[i]) + 1;
if (l == 0)
return;
if (l > 84)
errx(EX_DATAERR,
"comment too long (max 80 chars)");
l = 1 + (l+3)/4;
cmd->len = (cmd->len & (F_NOT | F_OR)) | l;
for (i = 0; av[i] != NULL; i++) {
strcpy(p, av[i]);
p += strlen(av[i]);
*p++ = ' ';
}
*(--p) = '\0';
}
/*
* A function to fill simple commands of size 1.
* Existing flags are preserved.
*/
static void
fill_cmd(ipfw_insn *cmd, enum ipfw_opcodes opcode, int flags, uint16_t arg)
{
cmd->opcode = opcode;
cmd->len = ((cmd->len | flags) & (F_NOT | F_OR)) | 1;
cmd->arg1 = arg;
}
/*
* Fetch and add the MAC address and type, with masks. This generates one or
* two microinstructions, and returns the pointer to the last one.
*/
static ipfw_insn *
add_mac(ipfw_insn *cmd, char *av[])
{
ipfw_insn_mac *mac;
if ( ( av[0] == NULL ) || ( av[1] == NULL ) )
errx(EX_DATAERR, "MAC dst src");
cmd->opcode = O_MACADDR2;
cmd->len = (cmd->len & (F_NOT | F_OR)) | F_INSN_SIZE(ipfw_insn_mac);
mac = (ipfw_insn_mac *)cmd;
get_mac_addr_mask(av[0], mac->addr, mac->mask); /* dst */
get_mac_addr_mask(av[1], &(mac->addr[ETHER_ADDR_LEN]),
&(mac->mask[ETHER_ADDR_LEN])); /* src */
return cmd;
}
static ipfw_insn *
add_mactype(ipfw_insn *cmd, char *av)
{
if (!av)
errx(EX_DATAERR, "missing MAC type");
if (strcmp(av, "any") != 0) { /* we have a non-null type */
fill_newports((ipfw_insn_u16 *)cmd, av, IPPROTO_ETHERTYPE);
cmd->opcode = O_MAC_TYPE;
return cmd;
} else
return NULL;
}
static ipfw_insn *
add_proto0(ipfw_insn *cmd, char *av, u_char *protop)
{
struct protoent *pe;
char *ep;
int proto;
proto = strtol(av, &ep, 10);
if (*ep != '\0' || proto <= 0) {
if ((pe = getprotobyname(av)) == NULL)
return NULL;
proto = pe->p_proto;
}
fill_cmd(cmd, O_PROTO, 0, proto);
*protop = proto;
return cmd;
}
static ipfw_insn *
add_proto(ipfw_insn *cmd, char *av, u_char *protop)
{
u_char proto = IPPROTO_IP;
if (_substrcmp(av, "all") == 0 || strcmp(av, "ip") == 0)
; /* do not set O_IP4 nor O_IP6 */
else if (strcmp(av, "ip4") == 0)
/* explicit "just IPv4" rule */
fill_cmd(cmd, O_IP4, 0, 0);
else if (strcmp(av, "ip6") == 0) {
/* explicit "just IPv6" rule */
proto = IPPROTO_IPV6;
fill_cmd(cmd, O_IP6, 0, 0);
} else
return add_proto0(cmd, av, protop);
*protop = proto;
return cmd;
}
static ipfw_insn *
add_proto_compat(ipfw_insn *cmd, char *av, u_char *protop)
{
u_char proto = IPPROTO_IP;
if (_substrcmp(av, "all") == 0 || strcmp(av, "ip") == 0)
; /* do not set O_IP4 nor O_IP6 */
else if (strcmp(av, "ipv4") == 0 || strcmp(av, "ip4") == 0)
/* explicit "just IPv4" rule */
fill_cmd(cmd, O_IP4, 0, 0);
else if (strcmp(av, "ipv6") == 0 || strcmp(av, "ip6") == 0) {
/* explicit "just IPv6" rule */
proto = IPPROTO_IPV6;
fill_cmd(cmd, O_IP6, 0, 0);
} else
return add_proto0(cmd, av, protop);
*protop = proto;
return cmd;
}
static ipfw_insn *
add_srcip(ipfw_insn *cmd, char *av)
{
fill_ip((ipfw_insn_ip *)cmd, av);
if (cmd->opcode == O_IP_DST_SET) /* set */
cmd->opcode = O_IP_SRC_SET;
else if (cmd->opcode == O_IP_DST_LOOKUP) /* table */
cmd->opcode = O_IP_SRC_LOOKUP;
else if (F_LEN(cmd) == F_INSN_SIZE(ipfw_insn)) /* me */
cmd->opcode = O_IP_SRC_ME;
else if (F_LEN(cmd) == F_INSN_SIZE(ipfw_insn_u32)) /* one IP */
cmd->opcode = O_IP_SRC;
else /* addr/mask */
cmd->opcode = O_IP_SRC_MASK;
return cmd;
}
static ipfw_insn *
add_dstip(ipfw_insn *cmd, char *av)
{
fill_ip((ipfw_insn_ip *)cmd, av);
if (cmd->opcode == O_IP_DST_SET) /* set */
;
else if (cmd->opcode == O_IP_DST_LOOKUP) /* table */
;
else if (F_LEN(cmd) == F_INSN_SIZE(ipfw_insn)) /* me */
cmd->opcode = O_IP_DST_ME;
else if (F_LEN(cmd) == F_INSN_SIZE(ipfw_insn_u32)) /* one IP */
cmd->opcode = O_IP_DST;
else /* addr/mask */
cmd->opcode = O_IP_DST_MASK;
return cmd;
}
static ipfw_insn *
add_ports(ipfw_insn *cmd, char *av, u_char proto, int opcode)
{
/* XXX "any" is trapped before. Perhaps "to" */
if (_substrcmp(av, "any") == 0) {
return NULL;
} else if (fill_newports((ipfw_insn_u16 *)cmd, av, proto)) {
/* XXX todo: check that we have a protocol with ports */
cmd->opcode = opcode;
return cmd;
}
return NULL;
}
static ipfw_insn *
add_src(ipfw_insn *cmd, char *av, u_char proto)
{
struct in6_addr a;
char *host, *ch;
ipfw_insn *ret = NULL;
if ((host = strdup(av)) == NULL)
return NULL;
if ((ch = strrchr(host, '/')) != NULL)
*ch = '\0';
if (proto == IPPROTO_IPV6 || strcmp(av, "me6") == 0 ||
inet_pton(AF_INET6, host, &a) == 1)
ret = add_srcip6(cmd, av);
/* XXX: should check for IPv4, not !IPv6 */
if (ret == NULL && (proto == IPPROTO_IP || strcmp(av, "me") == 0 ||
inet_pton(AF_INET6, host, &a) != 1))
ret = add_srcip(cmd, av);
if (ret == NULL && strcmp(av, "any") != 0)
ret = cmd;
free(host);
return ret;
}
static ipfw_insn *
add_dst(ipfw_insn *cmd, char *av, u_char proto)
{
struct in6_addr a;
char *host, *ch;
ipfw_insn *ret = NULL;
if ((host = strdup(av)) == NULL)
return NULL;
if ((ch = strrchr(host, '/')) != NULL)
*ch = '\0';
if (proto == IPPROTO_IPV6 || strcmp(av, "me6") == 0 ||
inet_pton(AF_INET6, host, &a) == 1)
ret = add_dstip6(cmd, av);
/* XXX: should check for IPv4, not !IPv6 */
if (ret == NULL && (proto == IPPROTO_IP || strcmp(av, "me") == 0 ||
inet_pton(AF_INET6, host, &a) != 1))
ret = add_dstip(cmd, av);
if (ret == NULL && strcmp(av, "any") != 0)
ret = cmd;
free(host);
return ret;
}
/*
* Parse arguments and assemble the microinstructions which make up a rule.
* Rules are added into the 'rulebuf' and then copied in the correct order
* into the actual rule.
*
* The syntax for a rule starts with the action, followed by
* optional action parameters, and the various match patterns.
* In the assembled microcode, the first opcode must be an O_PROBE_STATE
* (generated if the rule includes a keep-state option), then the
* various match patterns, log/altq actions, and the actual action.
*
*/
void
ipfw_add(char *av[])
{
/*
* rules are added into the 'rulebuf' and then copied in
* the correct order into the actual rule.
* Some things that need to go out of order (prob, action etc.)
* go into actbuf[].
*/
static uint32_t rulebuf[255], actbuf[255], cmdbuf[255];
ipfw_insn *src, *dst, *cmd, *action, *prev=NULL;
ipfw_insn *first_cmd; /* first match pattern */
struct ip_fw *rule;
/*
* various flags used to record that we entered some fields.
*/
ipfw_insn *have_state = NULL; /* check-state or keep-state */
ipfw_insn *have_log = NULL, *have_altq = NULL, *have_tag = NULL;
size_t len;
int i;
int open_par = 0; /* open parenthesis ( */
/* proto is here because it is used to fetch ports */
u_char proto = IPPROTO_IP; /* default protocol */
double match_prob = 1; /* match probability, default is always match */
bzero(actbuf, sizeof(actbuf)); /* actions go here */
bzero(cmdbuf, sizeof(cmdbuf));
bzero(rulebuf, sizeof(rulebuf));
rule = (struct ip_fw *)rulebuf;
cmd = (ipfw_insn *)cmdbuf;
action = (ipfw_insn *)actbuf;
av++;
/* [rule N] -- Rule number optional */
if (av[0] && isdigit(**av)) {
rule->rulenum = atoi(*av);
av++;
}
/* [set N] -- set number (0..RESVD_SET), optional */
if (av[0] && !av[1] && _substrcmp(*av, "set") == 0) {
int set = strtoul(av[1], NULL, 10);
if (set < 0 || set > RESVD_SET)
errx(EX_DATAERR, "illegal set %s", av[1]);
rule->set = set;
av += 2;
}
/* [prob D] -- match probability, optional */
if (av[0] && av[1] && _substrcmp(*av, "prob") == 0) {
match_prob = strtod(av[1], NULL);
if (match_prob <= 0 || match_prob > 1)
errx(EX_DATAERR, "illegal match prob. %s", av[1]);
av += 2;
}
/* action -- mandatory */
NEED1("missing action");
i = match_token(rule_actions, *av);
av++;
action->len = 1; /* default */
switch(i) {
case TOK_CHECKSTATE:
have_state = action;
action->opcode = O_CHECK_STATE;
break;
case TOK_ACCEPT:
action->opcode = O_ACCEPT;
break;
case TOK_DENY:
action->opcode = O_DENY;
action->arg1 = 0;
break;
case TOK_REJECT:
action->opcode = O_REJECT;
action->arg1 = ICMP_UNREACH_HOST;
break;
case TOK_RESET:
action->opcode = O_REJECT;
action->arg1 = ICMP_REJECT_RST;
break;
case TOK_RESET6:
action->opcode = O_UNREACH6;
action->arg1 = ICMP6_UNREACH_RST;
break;
case TOK_UNREACH:
action->opcode = O_REJECT;
NEED1("missing reject code");
fill_reject_code(&action->arg1, *av);
av++;
break;
case TOK_UNREACH6:
action->opcode = O_UNREACH6;
NEED1("missing unreach code");
fill_unreach6_code(&action->arg1, *av);
av++;
break;
case TOK_COUNT:
action->opcode = O_COUNT;
break;
case TOK_NAT:
action->opcode = O_NAT;
action->len = F_INSN_SIZE(ipfw_insn_nat);
goto chkarg;
case TOK_QUEUE:
action->opcode = O_QUEUE;
goto chkarg;
case TOK_PIPE:
action->opcode = O_PIPE;
goto chkarg;
case TOK_SKIPTO:
action->opcode = O_SKIPTO;
goto chkarg;
case TOK_NETGRAPH:
action->opcode = O_NETGRAPH;
goto chkarg;
case TOK_NGTEE:
action->opcode = O_NGTEE;
goto chkarg;
case TOK_DIVERT:
action->opcode = O_DIVERT;
goto chkarg;
case TOK_TEE:
action->opcode = O_TEE;
chkarg:
if (!av[0])
errx(EX_USAGE, "missing argument for %s", *(av - 1));
if (isdigit(**av)) {
action->arg1 = strtoul(*av, NULL, 10);
if (action->arg1 <= 0 || action->arg1 >= IP_FW_TABLEARG)
errx(EX_DATAERR, "illegal argument for %s",
*(av - 1));
} else if (_substrcmp(*av, "tablearg") == 0) {
action->arg1 = IP_FW_TABLEARG;
} else if (i == TOK_DIVERT || i == TOK_TEE) {
struct servent *s;
setservent(1);
s = getservbyname(av[0], "divert");
if (s != NULL)
action->arg1 = ntohs(s->s_port);
else
errx(EX_DATAERR, "illegal divert/tee port");
} else
errx(EX_DATAERR, "illegal argument for %s", *(av - 1));
av++;
break;
case TOK_FORWARD: {
ipfw_insn_sa *p = (ipfw_insn_sa *)action;
char *s, *end;
NEED1("missing forward address[:port]");
action->opcode = O_FORWARD_IP;
action->len = F_INSN_SIZE(ipfw_insn_sa);
/*
* In the kernel we assume AF_INET and use only
* sin_port and sin_addr. Remember to set sin_len as
* the routing code seems to use it too.
*/
p->sa.sin_family = AF_INET;
p->sa.sin_len = sizeof(struct sockaddr_in);
p->sa.sin_port = 0;
/*
* locate the address-port separator (':' or ',')
*/
s = strchr(*av, ':');
if (s == NULL)
s = strchr(*av, ',');
if (s != NULL) {
*(s++) = '\0';
i = strtoport(s, &end, 0 /* base */, 0 /* proto */);
if (s == end)
errx(EX_DATAERR,
"illegal forwarding port ``%s''", s);
p->sa.sin_port = (u_short)i;
}
if (_substrcmp(*av, "tablearg") == 0)
p->sa.sin_addr.s_addr = INADDR_ANY;
else
lookup_host(*av, &(p->sa.sin_addr));
av++;
break;
}
case TOK_COMMENT:
/* pretend it is a 'count' rule followed by the comment */
action->opcode = O_COUNT;
av--; /* go back... */
break;
case TOK_SETFIB:
{
int numfibs;
size_t intsize = sizeof(int);
action->opcode = O_SETFIB;
NEED1("missing fib number");
action->arg1 = strtoul(*av, NULL, 10);
if (sysctlbyname("net.fibs", &numfibs, &intsize, NULL, 0) == -1)
errx(EX_DATAERR, "fibs not suported.\n");
if (action->arg1 >= numfibs) /* Temporary */
errx(EX_DATAERR, "fib too large.\n");
av++;
break;
}
case TOK_REASS:
action->opcode = O_REASS;
break;
default:
errx(EX_DATAERR, "invalid action %s\n", av[-1]);
}
action = next_cmd(action);
/*
* [altq queuename] -- altq tag, optional
* [log [logamount N]] -- log, optional
*
* If they exist, it go first in the cmdbuf, but then it is
* skipped in the copy section to the end of the buffer.
*/
while (av[0] != NULL && (i = match_token(rule_action_params, *av)) != -1) {
av++;
switch (i) {
case TOK_LOG:
{
ipfw_insn_log *c = (ipfw_insn_log *)cmd;
int l;
if (have_log)
errx(EX_DATAERR,
"log cannot be specified more than once");
have_log = (ipfw_insn *)c;
cmd->len = F_INSN_SIZE(ipfw_insn_log);
cmd->opcode = O_LOG;
if (av[0] && _substrcmp(*av, "logamount") == 0) {
av++;
NEED1("logamount requires argument");
l = atoi(*av);
if (l < 0)
errx(EX_DATAERR,
"logamount must be positive");
c->max_log = l;
av++;
} else {
len = sizeof(c->max_log);
if (sysctlbyname("net.inet.ip.fw.verbose_limit",
&c->max_log, &len, NULL, 0) == -1)
errx(1, "sysctlbyname(\"%s\")",
"net.inet.ip.fw.verbose_limit");
}
}
break;
#ifndef NO_ALTQ
case TOK_ALTQ:
{
ipfw_insn_altq *a = (ipfw_insn_altq *)cmd;
NEED1("missing altq queue name");
if (have_altq)
errx(EX_DATAERR,
"altq cannot be specified more than once");
have_altq = (ipfw_insn *)a;
cmd->len = F_INSN_SIZE(ipfw_insn_altq);
cmd->opcode = O_ALTQ;
a->qid = altq_name_to_qid(*av);
av++;
}
break;
#endif
case TOK_TAG:
case TOK_UNTAG: {
uint16_t tag;
if (have_tag)
errx(EX_USAGE, "tag and untag cannot be "
"specified more than once");
GET_UINT_ARG(tag, IPFW_ARG_MIN, IPFW_ARG_MAX, i,
rule_action_params);
have_tag = cmd;
fill_cmd(cmd, O_TAG, (i == TOK_TAG) ? 0: F_NOT, tag);
av++;
break;
}
default:
abort();
}
cmd = next_cmd(cmd);
}
if (have_state) /* must be a check-state, we are done */
goto done;
#define OR_START(target) \
if (av[0] && (*av[0] == '(' || *av[0] == '{')) { \
if (open_par) \
errx(EX_USAGE, "nested \"(\" not allowed\n"); \
prev = NULL; \
open_par = 1; \
if ( (av[0])[1] == '\0') { \
av++; \
} else \
(*av)++; \
} \
target: \
#define CLOSE_PAR \
if (open_par) { \
if (av[0] && ( \
strcmp(*av, ")") == 0 || \
strcmp(*av, "}") == 0)) { \
prev = NULL; \
open_par = 0; \
av++; \
} else \
errx(EX_USAGE, "missing \")\"\n"); \
}
#define NOT_BLOCK \
if (av[0] && _substrcmp(*av, "not") == 0) { \
if (cmd->len & F_NOT) \
errx(EX_USAGE, "double \"not\" not allowed\n"); \
cmd->len |= F_NOT; \
av++; \
}
#define OR_BLOCK(target) \
if (av[0] && _substrcmp(*av, "or") == 0) { \
if (prev == NULL || open_par == 0) \
errx(EX_DATAERR, "invalid OR block"); \
prev->len |= F_OR; \
av++; \
goto target; \
} \
CLOSE_PAR;
first_cmd = cmd;
#if 0
/*
* MAC addresses, optional.
* If we have this, we skip the part "proto from src to dst"
* and jump straight to the option parsing.
*/
NOT_BLOCK;
NEED1("missing protocol");
if (_substrcmp(*av, "MAC") == 0 ||
_substrcmp(*av, "mac") == 0) {
av++; /* the "MAC" keyword */
add_mac(cmd, av); /* exits in case of errors */
cmd = next_cmd(cmd);
av += 2; /* dst-mac and src-mac */
NOT_BLOCK;
NEED1("missing mac type");
if (add_mactype(cmd, av[0]))
cmd = next_cmd(cmd);
av++; /* any or mac-type */
goto read_options;
}
#endif
/*
* protocol, mandatory
*/
OR_START(get_proto);
NOT_BLOCK;
NEED1("missing protocol");
if (add_proto_compat(cmd, *av, &proto)) {
av++;
if (F_LEN(cmd) != 0) {
prev = cmd;
cmd = next_cmd(cmd);
}
} else if (first_cmd != cmd) {
errx(EX_DATAERR, "invalid protocol ``%s''", *av);
} else
goto read_options;
OR_BLOCK(get_proto);
/*
* "from", mandatory
*/
if ((av[0] == NULL) || _substrcmp(*av, "from") != 0)
errx(EX_USAGE, "missing ``from''");
av++;
/*
* source IP, mandatory
*/
OR_START(source_ip);
NOT_BLOCK; /* optional "not" */
NEED1("missing source address");
if (add_src(cmd, *av, proto)) {
av++;
if (F_LEN(cmd) != 0) { /* ! any */
prev = cmd;
cmd = next_cmd(cmd);
}
} else
errx(EX_USAGE, "bad source address %s", *av);
OR_BLOCK(source_ip);
/*
* source ports, optional
*/
NOT_BLOCK; /* optional "not" */
if ( av[0] != NULL ) {
if (_substrcmp(*av, "any") == 0 ||
add_ports(cmd, *av, proto, O_IP_SRCPORT)) {
av++;
if (F_LEN(cmd) != 0)
cmd = next_cmd(cmd);
}
}
/*
* "to", mandatory
*/
if ( (av[0] == NULL) || _substrcmp(*av, "to") != 0 )
errx(EX_USAGE, "missing ``to''");
av++;
/*
* destination, mandatory
*/
OR_START(dest_ip);
NOT_BLOCK; /* optional "not" */
NEED1("missing dst address");
if (add_dst(cmd, *av, proto)) {
av++;
if (F_LEN(cmd) != 0) { /* ! any */
prev = cmd;
cmd = next_cmd(cmd);
}
} else
errx( EX_USAGE, "bad destination address %s", *av);
OR_BLOCK(dest_ip);
/*
* dest. ports, optional
*/
NOT_BLOCK; /* optional "not" */
if (av[0]) {
if (_substrcmp(*av, "any") == 0 ||
add_ports(cmd, *av, proto, O_IP_DSTPORT)) {
av++;
if (F_LEN(cmd) != 0)
cmd = next_cmd(cmd);
}
}
read_options:
if (av[0] && first_cmd == cmd) {
/*
* nothing specified so far, store in the rule to ease
* printout later.
*/
rule->_pad = 1;
}
prev = NULL;
while ( av[0] != NULL ) {
char *s;
ipfw_insn_u32 *cmd32; /* alias for cmd */
s = *av;
cmd32 = (ipfw_insn_u32 *)cmd;
if (*s == '!') { /* alternate syntax for NOT */
if (cmd->len & F_NOT)
errx(EX_USAGE, "double \"not\" not allowed\n");
cmd->len = F_NOT;
s++;
}
i = match_token(rule_options, s);
av++;
switch(i) {
case TOK_NOT:
if (cmd->len & F_NOT)
errx(EX_USAGE, "double \"not\" not allowed\n");
cmd->len = F_NOT;
break;
case TOK_OR:
if (open_par == 0 || prev == NULL)
errx(EX_USAGE, "invalid \"or\" block\n");
prev->len |= F_OR;
break;
case TOK_STARTBRACE:
if (open_par)
errx(EX_USAGE, "+nested \"(\" not allowed\n");
open_par = 1;
break;
case TOK_ENDBRACE:
if (!open_par)
errx(EX_USAGE, "+missing \")\"\n");
open_par = 0;
prev = NULL;
break;
case TOK_IN:
fill_cmd(cmd, O_IN, 0, 0);
break;
case TOK_OUT:
cmd->len ^= F_NOT; /* toggle F_NOT */
fill_cmd(cmd, O_IN, 0, 0);
break;
case TOK_DIVERTED:
fill_cmd(cmd, O_DIVERTED, 0, 3);
break;
case TOK_DIVERTEDLOOPBACK:
fill_cmd(cmd, O_DIVERTED, 0, 1);
break;
case TOK_DIVERTEDOUTPUT:
fill_cmd(cmd, O_DIVERTED, 0, 2);
break;
case TOK_FRAG:
fill_cmd(cmd, O_FRAG, 0, 0);
break;
case TOK_LAYER2:
fill_cmd(cmd, O_LAYER2, 0, 0);
break;
case TOK_XMIT:
case TOK_RECV:
case TOK_VIA:
NEED1("recv, xmit, via require interface name"
" or address");
fill_iface((ipfw_insn_if *)cmd, av[0]);
av++;
if (F_LEN(cmd) == 0) /* not a valid address */
break;
if (i == TOK_XMIT)
cmd->opcode = O_XMIT;
else if (i == TOK_RECV)
cmd->opcode = O_RECV;
else if (i == TOK_VIA)
cmd->opcode = O_VIA;
break;
case TOK_ICMPTYPES:
NEED1("icmptypes requires list of types");
fill_icmptypes((ipfw_insn_u32 *)cmd, *av);
av++;
break;
case TOK_ICMP6TYPES:
NEED1("icmptypes requires list of types");
fill_icmp6types((ipfw_insn_icmp6 *)cmd, *av);
av++;
break;
case TOK_IPTTL:
NEED1("ipttl requires TTL");
if (strpbrk(*av, "-,")) {
if (!add_ports(cmd, *av, 0, O_IPTTL))
errx(EX_DATAERR, "invalid ipttl %s", *av);
} else
fill_cmd(cmd, O_IPTTL, 0, strtoul(*av, NULL, 0));
av++;
break;
case TOK_IPID:
NEED1("ipid requires id");
if (strpbrk(*av, "-,")) {
if (!add_ports(cmd, *av, 0, O_IPID))
errx(EX_DATAERR, "invalid ipid %s", *av);
} else
fill_cmd(cmd, O_IPID, 0, strtoul(*av, NULL, 0));
av++;
break;
case TOK_IPLEN:
NEED1("iplen requires length");
if (strpbrk(*av, "-,")) {
if (!add_ports(cmd, *av, 0, O_IPLEN))
errx(EX_DATAERR, "invalid ip len %s", *av);
} else
fill_cmd(cmd, O_IPLEN, 0, strtoul(*av, NULL, 0));
av++;
break;
case TOK_IPVER:
NEED1("ipver requires version");
fill_cmd(cmd, O_IPVER, 0, strtoul(*av, NULL, 0));
av++;
break;
case TOK_IPPRECEDENCE:
NEED1("ipprecedence requires value");
fill_cmd(cmd, O_IPPRECEDENCE, 0,
(strtoul(*av, NULL, 0) & 7) << 5);
av++;
break;
case TOK_IPOPTS:
NEED1("missing argument for ipoptions");
fill_flags(cmd, O_IPOPT, f_ipopts, *av);
av++;
break;
case TOK_IPTOS:
NEED1("missing argument for iptos");
fill_flags(cmd, O_IPTOS, f_iptos, *av);
av++;
break;
case TOK_UID:
NEED1("uid requires argument");
{
char *end;
uid_t uid;
struct passwd *pwd;
cmd->opcode = O_UID;
uid = strtoul(*av, &end, 0);
pwd = (*end == '\0') ? getpwuid(uid) : getpwnam(*av);
if (pwd == NULL)
errx(EX_DATAERR, "uid \"%s\" nonexistent", *av);
cmd32->d[0] = pwd->pw_uid;
cmd->len |= F_INSN_SIZE(ipfw_insn_u32);
av++;
}
break;
case TOK_GID:
NEED1("gid requires argument");
{
char *end;
gid_t gid;
struct group *grp;
cmd->opcode = O_GID;
gid = strtoul(*av, &end, 0);
grp = (*end == '\0') ? getgrgid(gid) : getgrnam(*av);
if (grp == NULL)
errx(EX_DATAERR, "gid \"%s\" nonexistent", *av);
cmd32->d[0] = grp->gr_gid;
cmd->len |= F_INSN_SIZE(ipfw_insn_u32);
av++;
}
break;
case TOK_JAIL:
NEED1("jail requires argument");
{
char *end;
int jid;
cmd->opcode = O_JAIL;
jid = (int)strtol(*av, &end, 0);
if (jid < 0 || *end != '\0')
errx(EX_DATAERR, "jail requires prison ID");
cmd32->d[0] = (uint32_t)jid;
cmd->len |= F_INSN_SIZE(ipfw_insn_u32);
av++;
}
break;
case TOK_ESTAB:
fill_cmd(cmd, O_ESTAB, 0, 0);
break;
case TOK_SETUP:
fill_cmd(cmd, O_TCPFLAGS, 0,
(TH_SYN) | ( (TH_ACK) & 0xff) <<8 );
break;
case TOK_TCPDATALEN:
NEED1("tcpdatalen requires length");
if (strpbrk(*av, "-,")) {
if (!add_ports(cmd, *av, 0, O_TCPDATALEN))
errx(EX_DATAERR, "invalid tcpdata len %s", *av);
} else
fill_cmd(cmd, O_TCPDATALEN, 0,
strtoul(*av, NULL, 0));
av++;
break;
case TOK_TCPOPTS:
NEED1("missing argument for tcpoptions");
fill_flags(cmd, O_TCPOPTS, f_tcpopts, *av);
av++;
break;
case TOK_TCPSEQ:
case TOK_TCPACK:
NEED1("tcpseq/tcpack requires argument");
cmd->len = F_INSN_SIZE(ipfw_insn_u32);
cmd->opcode = (i == TOK_TCPSEQ) ? O_TCPSEQ : O_TCPACK;
cmd32->d[0] = htonl(strtoul(*av, NULL, 0));
av++;
break;
case TOK_TCPWIN:
NEED1("tcpwin requires length");
fill_cmd(cmd, O_TCPWIN, 0,
htons(strtoul(*av, NULL, 0)));
av++;
break;
case TOK_TCPFLAGS:
NEED1("missing argument for tcpflags");
cmd->opcode = O_TCPFLAGS;
fill_flags(cmd, O_TCPFLAGS, f_tcpflags, *av);
av++;
break;
case TOK_KEEPSTATE:
if (open_par)
errx(EX_USAGE, "keep-state cannot be part "
"of an or block");
if (have_state)
errx(EX_USAGE, "only one of keep-state "
"and limit is allowed");
have_state = cmd;
fill_cmd(cmd, O_KEEP_STATE, 0, 0);
break;
case TOK_LIMIT: {
ipfw_insn_limit *c = (ipfw_insn_limit *)cmd;
int val;
if (open_par)
errx(EX_USAGE,
"limit cannot be part of an or block");
if (have_state)
errx(EX_USAGE, "only one of keep-state and "
"limit is allowed");
have_state = cmd;
cmd->len = F_INSN_SIZE(ipfw_insn_limit);
cmd->opcode = O_LIMIT;
c->limit_mask = c->conn_limit = 0;
while ( av[0] != NULL ) {
if ((val = match_token(limit_masks, *av)) <= 0)
break;
c->limit_mask |= val;
av++;
}
if (c->limit_mask == 0)
errx(EX_USAGE, "limit: missing limit mask");
GET_UINT_ARG(c->conn_limit, IPFW_ARG_MIN, IPFW_ARG_MAX,
TOK_LIMIT, rule_options);
av++;
break;
}
case TOK_PROTO:
NEED1("missing protocol");
if (add_proto(cmd, *av, &proto)) {
av++;
} else
errx(EX_DATAERR, "invalid protocol ``%s''",
*av);
break;
case TOK_SRCIP:
NEED1("missing source IP");
if (add_srcip(cmd, *av)) {
av++;
}
break;
case TOK_DSTIP:
NEED1("missing destination IP");
if (add_dstip(cmd, *av)) {
av++;
}
break;
case TOK_SRCIP6:
NEED1("missing source IP6");
if (add_srcip6(cmd, *av)) {
av++;
}
break;
case TOK_DSTIP6:
NEED1("missing destination IP6");
if (add_dstip6(cmd, *av)) {
av++;
}
break;
case TOK_SRCPORT:
NEED1("missing source port");
if (_substrcmp(*av, "any") == 0 ||
add_ports(cmd, *av, proto, O_IP_SRCPORT)) {
av++;
} else
errx(EX_DATAERR, "invalid source port %s", *av);
break;
case TOK_DSTPORT:
NEED1("missing destination port");
if (_substrcmp(*av, "any") == 0 ||
add_ports(cmd, *av, proto, O_IP_DSTPORT)) {
av++;
} else
errx(EX_DATAERR, "invalid destination port %s",
*av);
break;
case TOK_MAC:
if (add_mac(cmd, av))
av += 2;
break;
case TOK_MACTYPE:
NEED1("missing mac type");
if (!add_mactype(cmd, *av))
errx(EX_DATAERR, "invalid mac type %s", *av);
av++;
break;
case TOK_VERREVPATH:
fill_cmd(cmd, O_VERREVPATH, 0, 0);
break;
case TOK_VERSRCREACH:
fill_cmd(cmd, O_VERSRCREACH, 0, 0);
break;
case TOK_ANTISPOOF:
fill_cmd(cmd, O_ANTISPOOF, 0, 0);
break;
case TOK_IPSEC:
fill_cmd(cmd, O_IPSEC, 0, 0);
break;
case TOK_IPV6:
fill_cmd(cmd, O_IP6, 0, 0);
break;
case TOK_IPV4:
fill_cmd(cmd, O_IP4, 0, 0);
break;
case TOK_EXT6HDR:
fill_ext6hdr( cmd, *av );
av++;
break;
case TOK_FLOWID:
if (proto != IPPROTO_IPV6 )
errx( EX_USAGE, "flow-id filter is active "
"only for ipv6 protocol\n");
fill_flow6( (ipfw_insn_u32 *) cmd, *av );
av++;
break;
case TOK_COMMENT:
fill_comment(cmd, av);
av[0]=NULL;
break;
case TOK_TAGGED:
if (av[0] && strpbrk(*av, "-,")) {
if (!add_ports(cmd, *av, 0, O_TAGGED))
errx(EX_DATAERR, "tagged: invalid tag"
" list: %s", *av);
}
else {
uint16_t tag;
GET_UINT_ARG(tag, IPFW_ARG_MIN, IPFW_ARG_MAX,
TOK_TAGGED, rule_options);
fill_cmd(cmd, O_TAGGED, 0, tag);
}
av++;
break;
case TOK_FIB:
NEED1("fib requires fib number");
fill_cmd(cmd, O_FIB, 0, strtoul(*av, NULL, 0));
av++;
break;
case TOK_LOOKUP: {
ipfw_insn_u32 *c = (ipfw_insn_u32 *)cmd;
char *p;
int j;
if (av[0] && av[1])
errx(EX_USAGE, "format: lookup argument tablenum");
cmd->opcode = O_IP_DST_LOOKUP;
cmd->len |= F_INSN_SIZE(ipfw_insn) + 2;
i = match_token(rule_options, *av);
for (j = 0; lookup_key[j] >= 0 ; j++) {
if (i == lookup_key[j])
break;
}
if (lookup_key[j] <= 0)
errx(EX_USAGE, "format: cannot lookup on %s", *av);
c->d[1] = j; // i converted to option
av++;
cmd->arg1 = strtoul(*av, &p, 0);
if (p && *p)
errx(EX_USAGE, "format: lookup argument tablenum");
av++;
}
break;
default:
errx(EX_USAGE, "unrecognised option [%d] %s\n", i, s);
}
if (F_LEN(cmd) > 0) { /* prepare to advance */
prev = cmd;
cmd = next_cmd(cmd);
}
}
done:
/*
* Now copy stuff into the rule.
* If we have a keep-state option, the first instruction
* must be a PROBE_STATE (which is generated here).
* If we have a LOG option, it was stored as the first command,
* and now must be moved to the top of the action part.
*/
dst = (ipfw_insn *)rule->cmd;
/*
* First thing to write into the command stream is the match probability.
*/
if (match_prob != 1) { /* 1 means always match */
dst->opcode = O_PROB;
dst->len = 2;
*((int32_t *)(dst+1)) = (int32_t)(match_prob * 0x7fffffff);
dst += dst->len;
}
/*
* generate O_PROBE_STATE if necessary
*/
if (have_state && have_state->opcode != O_CHECK_STATE) {
fill_cmd(dst, O_PROBE_STATE, 0, 0);
dst = next_cmd(dst);
}
/* copy all commands but O_LOG, O_KEEP_STATE, O_LIMIT, O_ALTQ, O_TAG */
for (src = (ipfw_insn *)cmdbuf; src != cmd; src += i) {
i = F_LEN(src);
switch (src->opcode) {
case O_LOG:
case O_KEEP_STATE:
case O_LIMIT:
case O_ALTQ:
case O_TAG:
break;
default:
bcopy(src, dst, i * sizeof(uint32_t));
dst += i;
}
}
/*
* put back the have_state command as last opcode
*/
if (have_state && have_state->opcode != O_CHECK_STATE) {
i = F_LEN(have_state);
bcopy(have_state, dst, i * sizeof(uint32_t));
dst += i;
}
/*
* start action section
*/
rule->act_ofs = dst - rule->cmd;
/* put back O_LOG, O_ALTQ, O_TAG if necessary */
if (have_log) {
i = F_LEN(have_log);
bcopy(have_log, dst, i * sizeof(uint32_t));
dst += i;
}
if (have_altq) {
i = F_LEN(have_altq);
bcopy(have_altq, dst, i * sizeof(uint32_t));
dst += i;
}
if (have_tag) {
i = F_LEN(have_tag);
bcopy(have_tag, dst, i * sizeof(uint32_t));
dst += i;
}
/*
* copy all other actions
*/
for (src = (ipfw_insn *)actbuf; src != action; src += i) {
i = F_LEN(src);
bcopy(src, dst, i * sizeof(uint32_t));
dst += i;
}
rule->cmd_len = (uint32_t *)dst - (uint32_t *)(rule->cmd);
i = (char *)dst - (char *)rule;
if (do_cmd(IP_FW_ADD, rule, (uintptr_t)&i) == -1)
err(EX_UNAVAILABLE, "getsockopt(%s)", "IP_FW_ADD");
if (!co.do_quiet)
show_ipfw(rule, 0, 0);
}
/*
* clear the counters or the log counters.
*/
void
ipfw_zero(int ac, char *av[], int optname /* 0 = IP_FW_ZERO, 1 = IP_FW_RESETLOG */)
{
uint32_t arg, saved_arg;
int failed = EX_OK;
char const *errstr;
char const *name = optname ? "RESETLOG" : "ZERO";
optname = optname ? IP_FW_RESETLOG : IP_FW_ZERO;
av++; ac--;
if (!ac) {
/* clear all entries */
if (do_cmd(optname, NULL, 0) < 0)
err(EX_UNAVAILABLE, "setsockopt(IP_FW_%s)", name);
if (!co.do_quiet)
printf("%s.\n", optname == IP_FW_ZERO ?
"Accounting cleared":"Logging counts reset");
return;
}
while (ac) {
/* Rule number */
if (isdigit(**av)) {
arg = strtonum(*av, 0, 0xffff, &errstr);
if (errstr)
errx(EX_DATAERR,
"invalid rule number %s\n", *av);
saved_arg = arg;
if (co.use_set)
arg |= (1 << 24) | ((co.use_set - 1) << 16);
av++;
ac--;
if (do_cmd(optname, &arg, sizeof(arg))) {
warn("rule %u: setsockopt(IP_FW_%s)",
saved_arg, name);
failed = EX_UNAVAILABLE;
} else if (!co.do_quiet)
printf("Entry %d %s.\n", saved_arg,
optname == IP_FW_ZERO ?
"cleared" : "logging count reset");
} else {
errx(EX_USAGE, "invalid rule number ``%s''", *av);
}
}
if (failed != EX_OK)
exit(failed);
}
void
ipfw_flush(int force)
{
int cmd = co.do_pipe ? IP_DUMMYNET_FLUSH : IP_FW_FLUSH;
if (!force && !co.do_quiet) { /* need to ask user */
int c;
printf("Are you sure? [yn] ");
fflush(stdout);
do {
c = toupper(getc(stdin));
while (c != '\n' && getc(stdin) != '\n')
if (feof(stdin))
return; /* and do not flush */
} while (c != 'Y' && c != 'N');
printf("\n");
if (c == 'N') /* user said no */
return;
}
if (co.do_pipe) {
dummynet_flush();
return;
}
/* `ipfw set N flush` - is the same that `ipfw delete set N` */
if (co.use_set) {
uint32_t arg = ((co.use_set - 1) & 0xffff) | (1 << 24);
if (do_cmd(IP_FW_DEL, &arg, sizeof(arg)) < 0)
err(EX_UNAVAILABLE, "setsockopt(IP_FW_DEL)");
} else if (do_cmd(cmd, NULL, 0) < 0)
err(EX_UNAVAILABLE, "setsockopt(IP_%s_FLUSH)",
co.do_pipe ? "DUMMYNET" : "FW");
if (!co.do_quiet)
printf("Flushed all %s.\n", co.do_pipe ? "pipes" : "rules");
}
static void table_list(ipfw_table_entry ent, int need_header);
/*
* This one handles all table-related commands
* ipfw table N add addr[/masklen] [value]
* ipfw table N delete addr[/masklen]
* ipfw table {N | all} flush
* ipfw table {N | all} list
*/
void
ipfw_table_handler(int ac, char *av[])
{
ipfw_table_entry ent;
int do_add;
int is_all;
size_t len;
char *p;
uint32_t a;
uint32_t tables_max;
len = sizeof(tables_max);
if (sysctlbyname("net.inet.ip.fw.tables_max", &tables_max, &len,
NULL, 0) == -1) {
#ifdef IPFW_TABLES_MAX
warn("Warn: Failed to get the max tables number via sysctl. "
"Using the compiled in defaults. \nThe reason was");
tables_max = IPFW_TABLES_MAX;
#else
errx(1, "Failed sysctlbyname(\"net.inet.ip.fw.tables_max\")");
#endif
}
ac--; av++;
if (ac && isdigit(**av)) {
ent.tbl = atoi(*av);
is_all = 0;
ac--; av++;
} else if (ac && _substrcmp(*av, "all") == 0) {
ent.tbl = 0;
is_all = 1;
ac--; av++;
} else
errx(EX_USAGE, "table number or 'all' keyword required");
if (ent.tbl >= tables_max)
errx(EX_USAGE, "The table number exceeds the maximum allowed "
"value (%d)", tables_max - 1);
NEED1("table needs command");
if (is_all && _substrcmp(*av, "list") != 0
&& _substrcmp(*av, "flush") != 0)
errx(EX_USAGE, "table number required");
if (_substrcmp(*av, "add") == 0 ||
_substrcmp(*av, "delete") == 0) {
do_add = **av == 'a';
ac--; av++;
if (!ac)
errx(EX_USAGE, "IP address required");
p = strchr(*av, '/');
if (p) {
*p++ = '\0';
ent.masklen = atoi(p);
if (ent.masklen > 32)
errx(EX_DATAERR, "bad width ``%s''", p);
} else
ent.masklen = 32;
if (lookup_host(*av, (struct in_addr *)&ent.addr) != 0)
errx(EX_NOHOST, "hostname ``%s'' unknown", *av);
ac--; av++;
if (do_add && ac) {
unsigned int tval;
/* isdigit is a bit of a hack here.. */
if (strchr(*av, (int)'.') == NULL && isdigit(**av)) {
ent.value = strtoul(*av, NULL, 0);
} else {
if (lookup_host(*av, (struct in_addr *)&tval) == 0) {
/* The value must be stored in host order *
* so that the values < 65k can be distinguished */
ent.value = ntohl(tval);
} else {
errx(EX_NOHOST, "hostname ``%s'' unknown", *av);
}
}
} else
ent.value = 0;
if (do_cmd(do_add ? IP_FW_TABLE_ADD : IP_FW_TABLE_DEL,
&ent, sizeof(ent)) < 0) {
/* If running silent, don't bomb out on these errors. */
if (!(co.do_quiet && (errno == (do_add ? EEXIST : ESRCH))))
err(EX_OSERR, "setsockopt(IP_FW_TABLE_%s)",
do_add ? "ADD" : "DEL");
/* In silent mode, react to a failed add by deleting */
if (do_add) {
do_cmd(IP_FW_TABLE_DEL, &ent, sizeof(ent));
if (do_cmd(IP_FW_TABLE_ADD,
&ent, sizeof(ent)) < 0)
err(EX_OSERR,
"setsockopt(IP_FW_TABLE_ADD)");
}
}
} else if (_substrcmp(*av, "flush") == 0) {
a = is_all ? tables_max : (uint32_t)(ent.tbl + 1);
do {
if (do_cmd(IP_FW_TABLE_FLUSH, &ent.tbl,
sizeof(ent.tbl)) < 0)
err(EX_OSERR, "setsockopt(IP_FW_TABLE_FLUSH)");
} while (++ent.tbl < a);
} else if (_substrcmp(*av, "list") == 0) {
a = is_all ? tables_max : (uint32_t)(ent.tbl + 1);
do {
table_list(ent, is_all);
} while (++ent.tbl < a);
} else
errx(EX_USAGE, "invalid table command %s", *av);
}
static void
table_list(ipfw_table_entry ent, int need_header)
{
ipfw_table *tbl;
socklen_t l;
uint32_t a;
a = ent.tbl;
l = sizeof(a);
if (do_cmd(IP_FW_TABLE_GETSIZE, &a, (uintptr_t)&l) < 0)
err(EX_OSERR, "getsockopt(IP_FW_TABLE_GETSIZE)");
/* If a is zero we have nothing to do, the table is empty. */
if (a == 0)
return;
l = sizeof(*tbl) + a * sizeof(ipfw_table_entry);
tbl = safe_calloc(1, l);
tbl->tbl = ent.tbl;
if (do_cmd(IP_FW_TABLE_LIST, tbl, (uintptr_t)&l) < 0)
err(EX_OSERR, "getsockopt(IP_FW_TABLE_LIST)");
if (tbl->cnt && need_header)
printf("---table(%d)---\n", tbl->tbl);
for (a = 0; a < tbl->cnt; a++) {
unsigned int tval;
tval = tbl->ent[a].value;
if (co.do_value_as_ip) {
char tbuf[128];
strncpy(tbuf, inet_ntoa(*(struct in_addr *)
&tbl->ent[a].addr), 127);
/* inet_ntoa expects network order */
tval = htonl(tval);
printf("%s/%u %s\n", tbuf, tbl->ent[a].masklen,
inet_ntoa(*(struct in_addr *)&tval));
} else {
printf("%s/%u %u\n",
inet_ntoa(*(struct in_addr *)&tbl->ent[a].addr),
tbl->ent[a].masklen, tval);
}
}
free(tbl);
}