freebsd-nq/sys/netinet6/ip6_fw.c
Jake Burkholder 740a1973a6 Change the way that the queue(3) structures are declared; don't assume that
the type argument to *_HEAD and *_ENTRY is a struct.

Suggested by:	phk
Reviewed by:	phk
Approved by:	mdodd
2000-05-23 20:41:01 +00:00

1174 lines
29 KiB
C

/*
* Copyright (c) 1993 Daniel Boulet
* Copyright (c) 1994 Ugen J.S.Antsilevich
* Copyright (c) 1996 Alex Nash
*
* 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.
*
* $Id: ip6_fw.c,v 1.7 1999/08/31 12:25:57 shin Exp $
* $FreeBSD$
*/
/*
* Implement IPv6 packet firewall
*/
#include "opt_ip6fw.h"
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/queue.h>
#include <sys/kernel.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/time.h>
#include <net/if.h>
#include <net/route.h>
#include <netinet/in_systm.h>
#include <netinet/in.h>
#include <netinet/ip.h>
#include <netinet/in_pcb.h>
#include <netinet6/in6_var.h>
#include <netinet6/ip6.h>
#include <netinet6/icmp6.h>
#include <netinet6/ip6_fw.h>
#include <netinet/ip_var.h>
#include <netinet/tcp.h>
#include <netinet/tcp_seq.h>
#include <netinet/tcp_timer.h>
#include <netinet/tcp_var.h>
#include <netinet/udp.h>
#include <sys/sysctl.h>
#include <net/net_osdep.h>
MALLOC_DEFINE(M_IP6FW, "Ip6Fw/Ip6Acct", "Ip6Fw/Ip6Acct chain's");
static int fw6_debug = 1;
#ifdef IPV6FIREWALL_VERBOSE
static int fw6_verbose = 1;
#else
static int fw6_verbose = 0;
#endif
#ifdef IPV6FIREWALL_VERBOSE_LIMIT
static int fw6_verbose_limit = IPV6FIREWALL_VERBOSE_LIMIT;
#else
static int fw6_verbose_limit = 0;
#endif
LIST_HEAD (ip6_fw_head, struct ip6_fw_chain) ip6_fw_chain;
#ifdef SYSCTL_NODE
SYSCTL_DECL(_net_inet6_ip6);
SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, debug, CTLFLAG_RW, &fw6_debug, 0, "");
SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, verbose, CTLFLAG_RW, &fw6_verbose, 0, "");
SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, &fw6_verbose_limit, 0, "");
#endif
#define dprintf(a) if (!fw6_debug); else printf a
#define print_ip6(a) printf("[%s]", ip6_sprintf(a))
#define dprint_ip6(a) if (!fw6_debug); else print_ip6(a)
static int add_entry6 __P((struct ip6_fw_head *chainptr, struct ip6_fw *frwl));
static int del_entry6 __P((struct ip6_fw_head *chainptr, u_short number));
static int zero_entry6 __P((struct mbuf *m));
static struct ip6_fw *check_ip6fw_struct __P((struct ip6_fw *m));
static struct ip6_fw *check_ip6fw_mbuf __P((struct mbuf *fw));
static int ip6opts_match __P((struct ip6_hdr **ip6, struct ip6_fw *f,
struct mbuf **m,
int *off, int *nxt, u_short *offset));
static int port_match6 __P((u_short *portptr, int nports, u_short port,
int range_flag));
static int tcp6flg_match __P((struct tcphdr *tcp6, struct ip6_fw *f));
static int icmp6type_match __P((struct icmp6_hdr * icmp, struct ip6_fw * f));
static void ip6fw_report __P((struct ip6_fw *f, struct ip6_hdr *ip6,
struct ifnet *rif, struct ifnet *oif, int off, int nxt));
static int ip6_fw_chk __P((struct ip6_hdr **pip6,
struct ifnet *oif, u_int16_t *cookie, struct mbuf **m));
static int ip6_fw_ctl __P((int stage, struct mbuf **mm));
static char err_prefix[] = "ip6_fw_ctl:";
/*
* Returns 1 if the port is matched by the vector, 0 otherwise
*/
static
__inline int
port_match6(u_short *portptr, int nports, u_short port, int range_flag)
{
if (!nports)
return 1;
if (range_flag) {
if (portptr[0] <= port && port <= portptr[1]) {
return 1;
}
nports -= 2;
portptr += 2;
}
while (nports-- > 0) {
if (*portptr++ == port) {
return 1;
}
}
return 0;
}
static int
tcp6flg_match(struct tcphdr *tcp6, struct ip6_fw *f)
{
u_char flg_set, flg_clr;
if ((f->fw_tcpf & IPV6_FW_TCPF_ESTAB) &&
(tcp6->th_flags & (IPV6_FW_TCPF_RST | IPV6_FW_TCPF_ACK)))
return 1;
flg_set = tcp6->th_flags & f->fw_tcpf;
flg_clr = tcp6->th_flags & f->fw_tcpnf;
if (flg_set != f->fw_tcpf)
return 0;
if (flg_clr)
return 0;
return 1;
}
static int
icmp6type_match(struct icmp6_hdr *icmp6, struct ip6_fw *f)
{
int type;
if (!(f->fw_flg & IPV6_FW_F_ICMPBIT))
return(1);
type = icmp6->icmp6_type;
/* check for matching type in the bitmap */
if (type < IPV6_FW_ICMPTYPES_DIM * sizeof(unsigned) * 8 &&
(f->fw_icmp6types[type / (sizeof(unsigned) * 8)] &
(1U << (type % (8 * sizeof(unsigned))))))
return(1);
return(0); /* no match */
}
static int
is_icmp6_query(struct ip6_hdr *ip6, int off)
{
const struct icmp6_hdr *icmp6;
int icmp6_type;
icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
icmp6_type = icmp6->icmp6_type;
if (icmp6_type == ICMP6_ECHO_REQUEST ||
icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
icmp6_type == ICMP6_WRUREQUEST ||
icmp6_type == ICMP6_FQDN_QUERY ||
icmp6_type == ICMP6_NI_QUERY)
return(1);
return(0);
}
static int
ip6opts_match(struct ip6_hdr **pip6, struct ip6_fw *f, struct mbuf **m,
int *off, int *nxt, u_short *offset)
{
int len;
struct ip6_hdr *ip6 = *pip6;
struct ip6_ext *ip6e;
u_char opts, nopts, nopts_sve;
opts = f->fw_ip6opt;
nopts = nopts_sve = f->fw_ip6nopt;
*nxt = ip6->ip6_nxt;
*off = sizeof(struct ip6_hdr);
len = ntohs(ip6->ip6_plen) + sizeof(struct ip6_hdr);
while (*off < len) {
ip6e = (struct ip6_ext *)((caddr_t) ip6 + *off);
if ((*m)->m_len < *off + sizeof(*ip6e))
goto opts_check; /* XXX */
switch(*nxt) {
case IPPROTO_FRAGMENT:
if ((*m)->m_len < *off + sizeof(struct ip6_frag)) {
struct ip6_frag *ip6f;
ip6f = (struct ip6_frag *) ((caddr_t)ip6 + *off);
*offset = ip6f->ip6f_offlg | IP6F_OFF_MASK;
}
opts &= ~IPV6_FW_IP6OPT_FRAG;
nopts &= ~IPV6_FW_IP6OPT_FRAG;
*off += sizeof(struct ip6_frag);
break;
case IPPROTO_AH:
opts &= ~IPV6_FW_IP6OPT_AH;
nopts &= ~IPV6_FW_IP6OPT_AH;
*off += (ip6e->ip6e_len + 2) << 2;
break;
default:
switch (*nxt) {
case IPPROTO_HOPOPTS:
opts &= ~IPV6_FW_IP6OPT_HOPOPT;
nopts &= ~IPV6_FW_IP6OPT_HOPOPT;
break;
case IPPROTO_ROUTING:
opts &= ~IPV6_FW_IP6OPT_ROUTE;
nopts &= ~IPV6_FW_IP6OPT_ROUTE;
break;
case IPPROTO_ESP:
opts &= ~IPV6_FW_IP6OPT_ESP;
nopts &= ~IPV6_FW_IP6OPT_ESP;
break;
case IPPROTO_NONE:
opts &= ~IPV6_FW_IP6OPT_NONXT;
nopts &= ~IPV6_FW_IP6OPT_NONXT;
goto opts_check;
break;
case IPPROTO_DSTOPTS:
opts &= ~IPV6_FW_IP6OPT_OPTS;
nopts &= ~IPV6_FW_IP6OPT_OPTS;
break;
default:
goto opts_check;
break;
}
*off += (ip6e->ip6e_len + 1) << 3;
break;
}
*nxt = ip6e->ip6e_nxt;
}
opts_check:
if (f->fw_ip6opt == f->fw_ip6nopt) /* XXX */
return 1;
if (opts == 0 && nopts == nopts_sve)
return 1;
else
return 0;
}
static
__inline int
iface_match(struct ifnet *ifp, union ip6_fw_if *ifu, int byname)
{
/* Check by name or by IP address */
if (byname) {
/* Check unit number (-1 is wildcard) */
if (ifu->fu_via_if.unit != -1
&& ifp->if_unit != ifu->fu_via_if.unit)
return(0);
/* Check name */
if (strncmp(ifp->if_name, ifu->fu_via_if.name, FW_IFNLEN))
return(0);
return(1);
} else if (!IN6_IS_ADDR_UNSPECIFIED(&ifu->fu_via_ip6)) { /* Zero == wildcard */
struct ifaddr *ia;
for (ia = ifp->if_addrlist.tqh_first; ia; ia = ia->ifa_list.tqe_next)
{
if (ia->ifa_addr == NULL)
continue;
if (ia->ifa_addr->sa_family != AF_INET6)
continue;
if (!IN6_ARE_ADDR_EQUAL(&ifu->fu_via_ip6,
&(((struct sockaddr_in6 *)
(ia->ifa_addr))->sin6_addr)))
continue;
return(1);
}
return(0);
}
return(1);
}
static void
ip6fw_report(struct ip6_fw *f, struct ip6_hdr *ip6,
struct ifnet *rif, struct ifnet *oif, int off, int nxt)
{
static int counter;
struct tcphdr *const tcp6 = (struct tcphdr *) ((caddr_t) ip6+ off);
struct udphdr *const udp = (struct udphdr *) ((caddr_t) ip6+ off);
struct icmp6_hdr *const icmp6 = (struct icmp6_hdr *) ((caddr_t) ip6+ off);
int count;
count = f ? f->fw_pcnt : ++counter;
if (fw6_verbose_limit != 0 && count > fw6_verbose_limit)
return;
/* Print command name */
printf("ip6fw: %d ", f ? f->fw_number : -1);
if (!f)
printf("Refuse");
else
switch (f->fw_flg & IPV6_FW_F_COMMAND) {
case IPV6_FW_F_DENY:
printf("Deny");
break;
case IPV6_FW_F_REJECT:
if (f->fw_reject_code == IPV6_FW_REJECT_RST)
printf("Reset");
else
printf("Unreach");
break;
case IPV6_FW_F_ACCEPT:
printf("Accept");
break;
case IPV6_FW_F_COUNT:
printf("Count");
break;
case IPV6_FW_F_DIVERT:
printf("Divert %d", f->fw_divert_port);
break;
case IPV6_FW_F_TEE:
printf("Tee %d", f->fw_divert_port);
break;
case IPV6_FW_F_SKIPTO:
printf("SkipTo %d", f->fw_skipto_rule);
break;
default:
printf("UNKNOWN");
break;
}
printf(" ");
switch (nxt) {
case IPPROTO_TCP:
printf("TCP ");
print_ip6(&ip6->ip6_src);
if (off > 0)
printf(":%d ", ntohs(tcp6->th_sport));
else
printf(" ");
print_ip6(&ip6->ip6_dst);
if (off > 0)
printf(":%d", ntohs(tcp6->th_dport));
break;
case IPPROTO_UDP:
printf("UDP ");
print_ip6(&ip6->ip6_src);
if (off > 0)
printf(":%d ", ntohs(udp->uh_sport));
else
printf(" ");
print_ip6(&ip6->ip6_dst);
if (off > 0)
printf(":%d", ntohs(udp->uh_dport));
break;
case IPPROTO_ICMPV6:
if (off > 0)
printf("IPV6-ICMP:%u.%u ", icmp6->icmp6_type, icmp6->icmp6_code);
else
printf("IPV6-ICMP ");
print_ip6(&ip6->ip6_src);
printf(" ");
print_ip6(&ip6->ip6_dst);
break;
default:
printf("P:%d ", nxt);
print_ip6(&ip6->ip6_src);
printf(" ");
print_ip6(&ip6->ip6_dst);
break;
}
if (oif)
printf(" out via %s", if_name(oif));
else if (rif)
printf(" in via %s", if_name(rif));
printf("\n");
if (fw6_verbose_limit != 0 && count == fw6_verbose_limit)
printf("ip6fw: limit reached on rule #%d\n",
f ? f->fw_number : -1);
}
/*
* Parameters:
*
* ip Pointer to packet header (struct ip6_hdr *)
* hlen Packet header length
* oif Outgoing interface, or NULL if packet is incoming
* #ifndef IP6FW_DIVERT_RESTART
* *cookie Ignore all divert/tee rules to this port (if non-zero)
* #else
* *cookie Skip up to the first rule past this rule number;
* #endif
* *m The packet; we set to NULL when/if we nuke it.
*
* Return value:
*
* 0 The packet is to be accepted and routed normally OR
* the packet was denied/rejected and has been dropped;
* in the latter case, *m is equal to NULL upon return.
* port Divert the packet to port.
*/
static int
ip6_fw_chk(struct ip6_hdr **pip6,
struct ifnet *oif, u_int16_t *cookie, struct mbuf **m)
{
struct ip6_fw_chain *chain;
struct ip6_fw *rule = NULL;
struct ip6_hdr *ip6 = *pip6;
struct ifnet *const rif = (*m)->m_pkthdr.rcvif;
u_short offset = 0;
int off = sizeof(struct ip6_hdr), nxt = ip6->ip6_nxt;
u_short src_port, dst_port;
#ifdef IP6FW_DIVERT_RESTART
u_int16_t skipto = *cookie;
#else
u_int16_t ignport = ntohs(*cookie);
#endif
*cookie = 0;
/*
* Go down the chain, looking for enlightment
* #ifdef IP6FW_DIVERT_RESTART
* If we've been asked to start at a given rule immediatly, do so.
* #endif
*/
chain = LIST_FIRST(&ip6_fw_chain);
#ifdef IP6FW_DIVERT_RESTART
if (skipto) {
if (skipto >= 65535)
goto dropit;
while (chain && (chain->rule->fw_number <= skipto)) {
chain = LIST_NEXT(chain, chain);
}
if (! chain) goto dropit;
}
#endif /* IP6FW_DIVERT_RESTART */
for (; chain; chain = LIST_NEXT(chain, chain)) {
register struct ip6_fw *const f = chain->rule;
if (oif) {
/* Check direction outbound */
if (!(f->fw_flg & IPV6_FW_F_OUT))
continue;
} else {
/* Check direction inbound */
if (!(f->fw_flg & IPV6_FW_F_IN))
continue;
}
#define IN6_ARE_ADDR_MASKEQUAL(x,y,z) (\
(((x)->s6_addr32[0] & (y)->s6_addr32[0]) == (z)->s6_addr32[0]) && \
(((x)->s6_addr32[1] & (y)->s6_addr32[1]) == (z)->s6_addr32[1]) && \
(((x)->s6_addr32[2] & (y)->s6_addr32[2]) == (z)->s6_addr32[2]) && \
(((x)->s6_addr32[3] & (y)->s6_addr32[3]) == (z)->s6_addr32[3]))
/* If src-addr doesn't match, not this rule. */
if (((f->fw_flg & IPV6_FW_F_INVSRC) != 0) ^
(!IN6_ARE_ADDR_MASKEQUAL(&ip6->ip6_src,&f->fw_smsk,&f->fw_src)))
continue;
/* If dest-addr doesn't match, not this rule. */
if (((f->fw_flg & IPV6_FW_F_INVDST) != 0) ^
(!IN6_ARE_ADDR_MASKEQUAL(&ip6->ip6_dst,&f->fw_dmsk,&f->fw_dst)))
continue;
#undef IN6_ARE_ADDR_MASKEQUAL
/* Interface check */
if ((f->fw_flg & IF6_FW_F_VIAHACK) == IF6_FW_F_VIAHACK) {
struct ifnet *const iface = oif ? oif : rif;
/* Backwards compatibility hack for "via" */
if (!iface || !iface_match(iface,
&f->fw_in_if, f->fw_flg & IPV6_FW_F_OIFNAME))
continue;
} else {
/* Check receive interface */
if ((f->fw_flg & IPV6_FW_F_IIFACE)
&& (!rif || !iface_match(rif,
&f->fw_in_if, f->fw_flg & IPV6_FW_F_IIFNAME)))
continue;
/* Check outgoing interface */
if ((f->fw_flg & IPV6_FW_F_OIFACE)
&& (!oif || !iface_match(oif,
&f->fw_out_if, f->fw_flg & IPV6_FW_F_OIFNAME)))
continue;
}
/* Check IP options */
if (!ip6opts_match(&ip6, f, m, &off, &nxt, &offset))
continue;
/* Fragments */
if ((f->fw_flg & IPV6_FW_F_FRAG) && !offset)
continue;
/* Check protocol; if wildcard, match */
if (f->fw_prot == IPPROTO_IPV6)
goto got_match;
/* If different, don't match */
if (nxt != f->fw_prot)
continue;
#define PULLUP_TO(len) do { \
if ((*m)->m_len < (len) \
&& (*m = m_pullup(*m, (len))) == 0) { \
goto dropit; \
} \
*pip6 = ip6 = mtod(*m, struct ip6_hdr *); \
} while (0)
/* Protocol specific checks */
switch (nxt) {
case IPPROTO_TCP:
{
struct tcphdr *tcp6;
if (offset == 1) { /* cf. RFC 1858 */
PULLUP_TO(off + 4); /* XXX ? */
goto bogusfrag;
}
if (offset != 0) {
/*
* TCP flags and ports aren't available in this
* packet -- if this rule specified either one,
* we consider the rule a non-match.
*/
if (f->fw_nports != 0 ||
f->fw_tcpf != f->fw_tcpnf)
continue;
break;
}
PULLUP_TO(off + 14);
tcp6 = (struct tcphdr *) ((caddr_t)ip6 + off);
if (f->fw_tcpf != f->fw_tcpnf && !tcp6flg_match(tcp6, f))
continue;
src_port = ntohs(tcp6->th_sport);
dst_port = ntohs(tcp6->th_dport);
goto check_ports;
}
case IPPROTO_UDP:
{
struct udphdr *udp;
if (offset != 0) {
/*
* Port specification is unavailable -- if this
* rule specifies a port, we consider the rule
* a non-match.
*/
if (f->fw_nports != 0)
continue;
break;
}
PULLUP_TO(off + 4);
udp = (struct udphdr *) ((caddr_t)ip6 + off);
src_port = ntohs(udp->uh_sport);
dst_port = ntohs(udp->uh_dport);
check_ports:
if (!port_match6(&f->fw_pts[0],
IPV6_FW_GETNSRCP(f), src_port,
f->fw_flg & IPV6_FW_F_SRNG))
continue;
if (!port_match6(&f->fw_pts[IPV6_FW_GETNSRCP(f)],
IPV6_FW_GETNDSTP(f), dst_port,
f->fw_flg & IPV6_FW_F_DRNG))
continue;
break;
}
case IPPROTO_ICMPV6:
{
struct icmp6_hdr *icmp;
if (offset != 0) /* Type isn't valid */
break;
PULLUP_TO(off + 2);
icmp = (struct icmp6_hdr *) ((caddr_t)ip6 + off);
if (!icmp6type_match(icmp, f))
continue;
break;
}
#undef PULLUP_TO
bogusfrag:
if (fw6_verbose)
ip6fw_report(NULL, ip6, rif, oif, off, nxt);
goto dropit;
}
got_match:
#ifndef IP6FW_DIVERT_RESTART
/* Ignore divert/tee rule if socket port is "ignport" */
switch (f->fw_flg & IPV6_FW_F_COMMAND) {
case IPV6_FW_F_DIVERT:
case IPV6_FW_F_TEE:
if (f->fw_divert_port == ignport)
continue; /* ignore this rule */
break;
}
#endif /* IP6FW_DIVERT_RESTART */
/* Update statistics */
f->fw_pcnt += 1;
f->fw_bcnt += ntohs(ip6->ip6_plen);
f->timestamp = time_second;
/* Log to console if desired */
if ((f->fw_flg & IPV6_FW_F_PRN) && fw6_verbose)
ip6fw_report(f, ip6, rif, oif, off, nxt);
/* Take appropriate action */
switch (f->fw_flg & IPV6_FW_F_COMMAND) {
case IPV6_FW_F_ACCEPT:
return(0);
case IPV6_FW_F_COUNT:
continue;
case IPV6_FW_F_DIVERT:
#ifdef IP6FW_DIVERT_RESTART
*cookie = f->fw_number;
#else
*cookie = htons(f->fw_divert_port);
#endif /* IP6FW_DIVERT_RESTART */
return(f->fw_divert_port);
case IPV6_FW_F_TEE:
/*
* XXX someday tee packet here, but beware that you
* can't use m_copym() or m_copypacket() because
* the divert input routine modifies the mbuf
* (and these routines only increment reference
* counts in the case of mbuf clusters), so need
* to write custom routine.
*/
continue;
case IPV6_FW_F_SKIPTO:
#ifdef DIAGNOSTIC
while (chain->chain.le_next
&& chain->chain.le_next->rule->fw_number
< f->fw_skipto_rule)
#else
while (chain->chain.le_next->rule->fw_number
< f->fw_skipto_rule)
#endif
chain = chain->chain.le_next;
continue;
}
/* Deny/reject this packet using this rule */
rule = f;
break;
}
#ifdef DIAGNOSTIC
/* Rule 65535 should always be there and should always match */
if (!chain)
panic("ip6_fw: chain");
#endif
/*
* At this point, we're going to drop the packet.
* Send a reject notice if all of the following are true:
*
* - The packet matched a reject rule
* - The packet is not an ICMP packet, or is an ICMP query packet
* - The packet is not a multicast or broadcast packet
*/
if ((rule->fw_flg & IPV6_FW_F_COMMAND) == IPV6_FW_F_REJECT
&& (nxt != IPPROTO_ICMPV6 || is_icmp6_query(ip6, off))
&& !((*m)->m_flags & (M_BCAST|M_MCAST))
&& !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
switch (rule->fw_reject_code) {
case IPV6_FW_REJECT_RST:
{
struct tcphdr *const tcp =
(struct tcphdr *) ((caddr_t)ip6 + off);
struct {
struct ip6_hdr ip6;
struct tcphdr th;
} ti;
tcp_seq ack, seq;
int flags;
if (offset != 0 || (tcp->th_flags & TH_RST))
break;
ti.ip6 = *ip6;
ti.th = *tcp;
NTOHL(ti.th.th_seq);
NTOHL(ti.th.th_ack);
ti.ip6.ip6_nxt = IPPROTO_TCP;
if (ti.th.th_flags & TH_ACK) {
ack = 0;
seq = ti.th.th_ack;
flags = TH_RST;
} else {
ack = ti.th.th_seq;
if (((*m)->m_flags & M_PKTHDR) != 0) {
ack += (*m)->m_pkthdr.len - off
- (ti.th.th_off << 2);
} else if (ip6->ip6_plen) {
ack += ntohs(ip6->ip6_plen) + sizeof(*ip6)
- off - (ti.th.th_off << 2);
} else {
m_freem(*m);
*m = 0;
break;
}
seq = 0;
flags = TH_RST|TH_ACK;
}
bcopy(&ti, ip6, sizeof(ti));
m_freem(*m);
*m = NULL;
break;
}
default: /* Send an ICMP unreachable using code */
if (oif)
(*m)->m_pkthdr.rcvif = oif;
icmp6_error(*m, ICMP6_DST_UNREACH,
rule->fw_reject_code, 0);
*m = NULL;
break;
}
}
dropit:
/*
* Finally, drop the packet.
*/
if (*m) {
m_freem(*m);
*m = NULL;
}
return(0);
}
static int
add_entry6(struct ip6_fw_head *chainptr, struct ip6_fw *frwl)
{
struct ip6_fw *ftmp = 0;
struct ip6_fw_chain *fwc = 0, *fcp, *fcpl = 0;
u_short nbr = 0;
int s;
fwc = malloc(sizeof *fwc, M_IP6FW, M_DONTWAIT);
ftmp = malloc(sizeof *ftmp, M_IP6FW, M_DONTWAIT);
if (!fwc || !ftmp) {
dprintf(("%s malloc said no\n", err_prefix));
if (fwc) free(fwc, M_IP6FW);
if (ftmp) free(ftmp, M_IP6FW);
return (ENOSPC);
}
bcopy(frwl, ftmp, sizeof(struct ip6_fw));
ftmp->fw_in_if.fu_via_if.name[FW_IFNLEN - 1] = '\0';
ftmp->fw_pcnt = 0L;
ftmp->fw_bcnt = 0L;
fwc->rule = ftmp;
s = splnet();
if (!chainptr->lh_first) {
LIST_INSERT_HEAD(chainptr, fwc, chain);
splx(s);
return(0);
} else if (ftmp->fw_number == (u_short)-1) {
if (fwc) free(fwc, M_IP6FW);
if (ftmp) free(ftmp, M_IP6FW);
splx(s);
dprintf(("%s bad rule number\n", err_prefix));
return (EINVAL);
}
/* If entry number is 0, find highest numbered rule and add 100 */
if (ftmp->fw_number == 0) {
for (fcp = chainptr->lh_first; fcp; fcp = fcp->chain.le_next) {
if (fcp->rule->fw_number != (u_short)-1)
nbr = fcp->rule->fw_number;
else
break;
}
if (nbr < (u_short)-1 - 100)
nbr += 100;
ftmp->fw_number = nbr;
}
/* Got a valid number; now insert it, keeping the list ordered */
for (fcp = chainptr->lh_first; fcp; fcp = fcp->chain.le_next) {
if (fcp->rule->fw_number > ftmp->fw_number) {
if (fcpl) {
LIST_INSERT_AFTER(fcpl, fwc, chain);
} else {
LIST_INSERT_HEAD(chainptr, fwc, chain);
}
break;
} else {
fcpl = fcp;
}
}
splx(s);
return (0);
}
static int
del_entry6(struct ip6_fw_head *chainptr, u_short number)
{
struct ip6_fw_chain *fcp;
int s;
s = splnet();
fcp = chainptr->lh_first;
if (number != (u_short)-1) {
for (; fcp; fcp = fcp->chain.le_next) {
if (fcp->rule->fw_number == number) {
LIST_REMOVE(fcp, chain);
splx(s);
free(fcp->rule, M_IP6FW);
free(fcp, M_IP6FW);
return 0;
}
}
}
splx(s);
return (EINVAL);
}
static int
zero_entry6(struct mbuf *m)
{
struct ip6_fw *frwl;
struct ip6_fw_chain *fcp;
int s;
if (m) {
if (m->m_len != sizeof(struct ip6_fw))
return(EINVAL);
frwl = mtod(m, struct ip6_fw *);
}
else
frwl = NULL;
/*
* It's possible to insert multiple chain entries with the
* same number, so we don't stop after finding the first
* match if zeroing a specific entry.
*/
s = splnet();
for (fcp = ip6_fw_chain.lh_first; fcp; fcp = fcp->chain.le_next)
if (!frwl || frwl->fw_number == fcp->rule->fw_number) {
fcp->rule->fw_bcnt = fcp->rule->fw_pcnt = 0;
fcp->rule->timestamp = 0;
}
splx(s);
if (fw6_verbose) {
if (frwl)
printf("ip6fw: Entry %d cleared.\n", frwl->fw_number);
else
printf("ip6fw: Accounting cleared.\n");
}
return(0);
}
static struct ip6_fw *
check_ip6fw_mbuf(struct mbuf *m)
{
/* Check length */
if (m->m_len != sizeof(struct ip6_fw)) {
dprintf(("%s len=%d, want %d\n", err_prefix, m->m_len,
sizeof(struct ip6_fw)));
return (NULL);
}
return(check_ip6fw_struct(mtod(m, struct ip6_fw *)));
}
static struct ip6_fw *
check_ip6fw_struct(struct ip6_fw *frwl)
{
/* Check for invalid flag bits */
if ((frwl->fw_flg & ~IPV6_FW_F_MASK) != 0) {
dprintf(("%s undefined flag bits set (flags=%x)\n",
err_prefix, frwl->fw_flg));
return (NULL);
}
/* Must apply to incoming or outgoing (or both) */
if (!(frwl->fw_flg & (IPV6_FW_F_IN | IPV6_FW_F_OUT))) {
dprintf(("%s neither in nor out\n", err_prefix));
return (NULL);
}
/* Empty interface name is no good */
if (((frwl->fw_flg & IPV6_FW_F_IIFNAME)
&& !*frwl->fw_in_if.fu_via_if.name)
|| ((frwl->fw_flg & IPV6_FW_F_OIFNAME)
&& !*frwl->fw_out_if.fu_via_if.name)) {
dprintf(("%s empty interface name\n", err_prefix));
return (NULL);
}
/* Sanity check interface matching */
if ((frwl->fw_flg & IF6_FW_F_VIAHACK) == IF6_FW_F_VIAHACK) {
; /* allow "via" backwards compatibility */
} else if ((frwl->fw_flg & IPV6_FW_F_IN)
&& (frwl->fw_flg & IPV6_FW_F_OIFACE)) {
dprintf(("%s outgoing interface check on incoming\n",
err_prefix));
return (NULL);
}
/* Sanity check port ranges */
if ((frwl->fw_flg & IPV6_FW_F_SRNG) && IPV6_FW_GETNSRCP(frwl) < 2) {
dprintf(("%s src range set but n_src_p=%d\n",
err_prefix, IPV6_FW_GETNSRCP(frwl)));
return (NULL);
}
if ((frwl->fw_flg & IPV6_FW_F_DRNG) && IPV6_FW_GETNDSTP(frwl) < 2) {
dprintf(("%s dst range set but n_dst_p=%d\n",
err_prefix, IPV6_FW_GETNDSTP(frwl)));
return (NULL);
}
if (IPV6_FW_GETNSRCP(frwl) + IPV6_FW_GETNDSTP(frwl) > IPV6_FW_MAX_PORTS) {
dprintf(("%s too many ports (%d+%d)\n",
err_prefix, IPV6_FW_GETNSRCP(frwl), IPV6_FW_GETNDSTP(frwl)));
return (NULL);
}
/*
* Protocols other than TCP/UDP don't use port range
*/
if ((frwl->fw_prot != IPPROTO_TCP) &&
(frwl->fw_prot != IPPROTO_UDP) &&
(IPV6_FW_GETNSRCP(frwl) || IPV6_FW_GETNDSTP(frwl))) {
dprintf(("%s port(s) specified for non TCP/UDP rule\n",
err_prefix));
return(NULL);
}
/*
* Rather than modify the entry to make such entries work,
* we reject this rule and require user level utilities
* to enforce whatever policy they deem appropriate.
*/
if ((frwl->fw_src.s6_addr32[0] & (~frwl->fw_smsk.s6_addr32[0])) ||
(frwl->fw_src.s6_addr32[1] & (~frwl->fw_smsk.s6_addr32[1])) ||
(frwl->fw_src.s6_addr32[2] & (~frwl->fw_smsk.s6_addr32[2])) ||
(frwl->fw_src.s6_addr32[3] & (~frwl->fw_smsk.s6_addr32[3])) ||
(frwl->fw_dst.s6_addr32[0] & (~frwl->fw_dmsk.s6_addr32[0])) ||
(frwl->fw_dst.s6_addr32[1] & (~frwl->fw_dmsk.s6_addr32[1])) ||
(frwl->fw_dst.s6_addr32[2] & (~frwl->fw_dmsk.s6_addr32[2])) ||
(frwl->fw_dst.s6_addr32[3] & (~frwl->fw_dmsk.s6_addr32[3]))) {
dprintf(("%s rule never matches\n", err_prefix));
return(NULL);
}
if ((frwl->fw_flg & IPV6_FW_F_FRAG) &&
(frwl->fw_prot == IPPROTO_UDP || frwl->fw_prot == IPPROTO_TCP)) {
if (frwl->fw_nports) {
dprintf(("%s cannot mix 'frag' and ports\n", err_prefix));
return(NULL);
}
if (frwl->fw_prot == IPPROTO_TCP &&
frwl->fw_tcpf != frwl->fw_tcpnf) {
dprintf(("%s cannot mix 'frag' with TCP flags\n", err_prefix));
return(NULL);
}
}
/* Check command specific stuff */
switch (frwl->fw_flg & IPV6_FW_F_COMMAND)
{
case IPV6_FW_F_REJECT:
if (frwl->fw_reject_code >= 0x100
&& !(frwl->fw_prot == IPPROTO_TCP
&& frwl->fw_reject_code == IPV6_FW_REJECT_RST)) {
dprintf(("%s unknown reject code\n", err_prefix));
return(NULL);
}
break;
case IPV6_FW_F_DIVERT: /* Diverting to port zero is invalid */
case IPV6_FW_F_TEE:
if (frwl->fw_divert_port == 0) {
dprintf(("%s can't divert to port 0\n", err_prefix));
return (NULL);
}
break;
case IPV6_FW_F_DENY:
case IPV6_FW_F_ACCEPT:
case IPV6_FW_F_COUNT:
case IPV6_FW_F_SKIPTO:
break;
default:
dprintf(("%s invalid command\n", err_prefix));
return(NULL);
}
return frwl;
}
static int
ip6_fw_ctl(int stage, struct mbuf **mm)
{
int error;
struct mbuf *m;
if (stage == IPV6_FW_GET) {
struct ip6_fw_chain *fcp = ip6_fw_chain.lh_first;
*mm = m = m_get(M_WAIT, MT_DATA); /* XXX */
if (!m)
return(ENOBUFS);
if (sizeof *(fcp->rule) > MLEN) {
MCLGET(m, M_WAIT);
if ((m->m_flags & M_EXT) == 0) {
m_free(m);
return(ENOBUFS);
}
}
for (; fcp; fcp = fcp->chain.le_next) {
memcpy(m->m_data, fcp->rule, sizeof *(fcp->rule));
m->m_len = sizeof *(fcp->rule);
m->m_next = m_get(M_WAIT, MT_DATA); /* XXX */
if (!m->m_next) {
m_freem(*mm);
return(ENOBUFS);
}
m = m->m_next;
if (sizeof *(fcp->rule) > MLEN) {
MCLGET(m, M_WAIT);
if ((m->m_flags & M_EXT) == 0) {
m_freem(*mm);
return(ENOBUFS);
}
}
m->m_len = 0;
}
return (0);
}
m = *mm;
/* only allow get calls if secure mode > 2 */
if (securelevel > 2) {
if (m) {
(void)m_freem(m);
*mm = 0;
}
return(EPERM);
}
if (stage == IPV6_FW_FLUSH) {
while (ip6_fw_chain.lh_first != NULL &&
ip6_fw_chain.lh_first->rule->fw_number != (u_short)-1) {
struct ip6_fw_chain *fcp = ip6_fw_chain.lh_first;
int s = splnet();
LIST_REMOVE(ip6_fw_chain.lh_first, chain);
splx(s);
free(fcp->rule, M_IP6FW);
free(fcp, M_IP6FW);
}
if (m) {
(void)m_freem(m);
*mm = 0;
}
return (0);
}
if (stage == IPV6_FW_ZERO) {
error = zero_entry6(m);
if (m) {
(void)m_freem(m);
*mm = 0;
}
return (error);
}
if (m == NULL) {
printf("%s NULL mbuf ptr\n", err_prefix);
return (EINVAL);
}
if (stage == IPV6_FW_ADD) {
struct ip6_fw *frwl = check_ip6fw_mbuf(m);
if (!frwl)
error = EINVAL;
else
error = add_entry6(&ip6_fw_chain, frwl);
if (m) {
(void)m_freem(m);
*mm = 0;
}
return error;
}
if (stage == IPV6_FW_DEL) {
if (m->m_len != sizeof(struct ip6_fw)) {
dprintf(("%s len=%d, want %d\n", err_prefix, m->m_len,
sizeof(struct ip6_fw)));
error = EINVAL;
} else if (mtod(m, struct ip6_fw *)->fw_number == (u_short)-1) {
dprintf(("%s can't delete rule 65535\n", err_prefix));
error = EINVAL;
} else
error = del_entry6(&ip6_fw_chain,
mtod(m, struct ip6_fw *)->fw_number);
if (m) {
(void)m_freem(m);
*mm = 0;
}
return error;
}
dprintf(("%s unknown request %d\n", err_prefix, stage));
if (m) {
(void)m_freem(m);
*mm = 0;
}
return (EINVAL);
}
void
ip6_fw_init(void)
{
struct ip6_fw default_rule;
ip6_fw_chk_ptr = ip6_fw_chk;
ip6_fw_ctl_ptr = ip6_fw_ctl;
LIST_INIT(&ip6_fw_chain);
bzero(&default_rule, sizeof default_rule);
default_rule.fw_prot = IPPROTO_IPV6;
default_rule.fw_number = (u_short)-1;
#ifdef IPV6FIREWALL_DEFAULT_TO_ACCEPT
default_rule.fw_flg |= IPV6_FW_F_ACCEPT;
#else
default_rule.fw_flg |= IPV6_FW_F_DENY;
#endif
default_rule.fw_flg |= IPV6_FW_F_IN | IPV6_FW_F_OUT;
if (check_ip6fw_struct(&default_rule) == NULL ||
add_entry6(&ip6_fw_chain, &default_rule))
panic(__FUNCTION__);
printf("IPv6 packet filtering initialized, ");
#ifdef IPV6FIREWALL_DEFAULT_TO_ACCEPT
printf("default to accept, ");
#endif
#ifndef IPV6FIREWALL_VERBOSE
printf("logging disabled\n");
#else
if (fw6_verbose_limit == 0)
printf("unlimited logging\n");
else
printf("logging limited to %d packets/entry\n",
fw6_verbose_limit);
#endif
}