NetBSD/dist/ipf/ip_fil.c

877 lines
18 KiB
C

/* $NetBSD: ip_fil.c,v 1.19 2010/06/01 08:53:20 plunky Exp $ */
/*
* Copyright (C) 1993-2001 by Darren Reed.
*
* See the IPFILTER.LICENCE file for details on licencing.
*/
#if !defined(lint)
static const char sccsid[] = "@(#)ip_fil.c 2.41 6/5/96 (C) 1993-2000 Darren Reed";
static const char rcsid[] = "@(#)Id: ip_fil.c,v 2.133.2.21 2009/12/27 06:55:08 darrenr Exp";
#endif
#ifndef SOLARIS
#define SOLARIS (defined(sun) && (defined(__svr4__) || defined(__SVR4)))
#endif
#include <sys/param.h>
#if defined(__FreeBSD__) && !defined(__FreeBSD_version)
# if defined(IPFILTER_LKM)
# ifndef __FreeBSD_cc_version
# include <osreldate.h>
# else
# if __FreeBSD_cc_version < 430000
# include <osreldate.h>
# endif
# endif
# endif
#endif
#include <sys/errno.h>
#if defined(__hpux) && (HPUXREV >= 1111) && !defined(_KERNEL)
# include <sys/kern_svcs.h>
#endif
#include <sys/types.h>
#define _KERNEL
#define KERNEL
#ifdef __OpenBSD__
struct file;
#endif
#include <sys/uio.h>
#undef _KERNEL
#undef KERNEL
#include <sys/file.h>
#include <sys/ioctl.h>
#ifdef __sgi
# include <sys/ptimers.h>
#endif
#include <sys/time.h>
#if !SOLARIS
# if (NetBSD > 199609) || (OpenBSD > 199603) || (__FreeBSD_version >= 300000)
# include <sys/dirent.h>
# else
# include <sys/dir.h>
# endif
#else
# include <sys/filio.h>
#endif
#ifndef linux
# include <sys/protosw.h>
#endif
#include <sys/socket.h>
#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#include <ctype.h>
#include <fcntl.h>
#ifdef __hpux
# define _NET_ROUTE_INCLUDED
#endif
#include <net/if.h>
#ifdef sun
# include <net/af.h>
#endif
#if __FreeBSD_version >= 300000
# include <net/if_var.h>
#endif
#ifdef __sgi
#include <sys/debug.h>
# ifdef IFF_DRVRLOCK /* IRIX6 */
#include <sys/hashing.h>
# endif
#endif
#if defined(__FreeBSD__) || defined(SOLARIS2)
# include "radix_ipf.h"
#endif
#ifndef __osf__
# include <net/route.h>
#endif
#include <netinet/in.h>
#if !(defined(__sgi) && !defined(IFF_DRVRLOCK)) /* IRIX < 6 */ && \
!defined(__hpux) && !defined(linux)
# include <netinet/in_var.h>
#endif
#include <netinet/in_systm.h>
#include <netinet/ip.h>
#if !defined(linux)
# include <netinet/ip_var.h>
#endif
#include <netinet/tcp.h>
#if defined(__osf__)
# include <netinet/tcp_timer.h>
#endif
#if defined(__osf__) || defined(__hpux) || defined(__sgi)
# include "radix_ipf_local.h"
# define _RADIX_H_
#endif
#include <netinet/udp.h>
#include <netinet/tcpip.h>
#include <netinet/ip_icmp.h>
#include <unistd.h>
#include <syslog.h>
#include <arpa/inet.h>
#ifdef __hpux
# undef _NET_ROUTE_INCLUDED
#endif
#include "netinet/ip_compat.h"
#include "netinet/ip_fil.h"
#include "netinet/ip_nat.h"
#include "netinet/ip_frag.h"
#include "netinet/ip_state.h"
#include "netinet/ip_proxy.h"
#include "netinet/ip_auth.h"
#ifdef IPFILTER_SYNC
#include "netinet/ip_sync.h"
#endif
#ifdef IPFILTER_SCAN
#include "netinet/ip_scan.h"
#endif
#include "netinet/ip_pool.h"
#ifdef IPFILTER_COMPILED
# include "netinet/ip_rules.h"
#endif
#if defined(__FreeBSD_version) && (__FreeBSD_version >= 300000)
# include <sys/malloc.h>
#endif
#ifdef __hpux
struct rtentry;
#endif
#include "md5.h"
#if !defined(__osf__) && !defined(__linux__)
extern struct protosw inetsw[];
#endif
#include "ipt.h"
static struct ifnet **ifneta = NULL;
static int nifs = 0;
static void fr_setifpaddr __P((struct ifnet *, char *));
void init_ifp __P((void));
#if defined(__sgi) && (IRIX < 60500)
static int no_output __P((struct ifnet *, struct mbuf *,
struct sockaddr *));
static int write_output __P((struct ifnet *, struct mbuf *,
struct sockaddr *));
#else
# if TRU64 >= 1885
static int no_output __P((struct ifnet *, struct mbuf *,
struct sockaddr *, struct rtentry *, char *));
static int write_output __P((struct ifnet *, struct mbuf *,
struct sockaddr *, struct rtentry *, char *));
# else
#if defined(__NetBSD_Version__) && (__NetBSD_Version__ >= 499001100)
static int no_output(struct ifnet *, struct mbuf *,
const struct sockaddr *, struct rtentry *);
static int write_output(struct ifnet *, struct mbuf *,
const struct sockaddr *, struct rtentry *);
#else
static int no_output __P((struct ifnet *, struct mbuf *,
struct sockaddr *, struct rtentry *));
static int write_output __P((struct ifnet *, struct mbuf *,
struct sockaddr *, struct rtentry *));
#endif
# endif
#endif
int ipfattach()
{
fr_running = 1;
return 0;
}
int ipfdetach()
{
fr_running = -1;
return 0;
}
/*
* Filter ioctl interface.
*/
int iplioctl(dev, cmd, data, mode)
int dev;
ioctlcmd_t cmd;
void *data;
int mode;
{
int error = 0, unit = 0, uid;
SPL_INT(s);
uid = getuid();
unit = dev;
SPL_NET(s);
error = fr_ioctlswitch(unit, data, cmd, mode, uid, NULL);
if (error != -1) {
SPL_X(s);
return error;
}
SPL_X(s);
return error;
}
void fr_forgetifp(ifp)
void *ifp;
{
register frentry_t *f;
WRITE_ENTER(&ipf_mutex);
for (f = ipacct[0][fr_active]; (f != NULL); f = f->fr_next)
if (f->fr_ifa == ifp)
f->fr_ifa = (void *)-1;
for (f = ipacct[1][fr_active]; (f != NULL); f = f->fr_next)
if (f->fr_ifa == ifp)
f->fr_ifa = (void *)-1;
for (f = ipfilter[0][fr_active]; (f != NULL); f = f->fr_next)
if (f->fr_ifa == ifp)
f->fr_ifa = (void *)-1;
for (f = ipfilter[1][fr_active]; (f != NULL); f = f->fr_next)
if (f->fr_ifa == ifp)
f->fr_ifa = (void *)-1;
#ifdef USE_INET6
for (f = ipacct6[0][fr_active]; (f != NULL); f = f->fr_next)
if (f->fr_ifa == ifp)
f->fr_ifa = (void *)-1;
for (f = ipacct6[1][fr_active]; (f != NULL); f = f->fr_next)
if (f->fr_ifa == ifp)
f->fr_ifa = (void *)-1;
for (f = ipfilter6[0][fr_active]; (f != NULL); f = f->fr_next)
if (f->fr_ifa == ifp)
f->fr_ifa = (void *)-1;
for (f = ipfilter6[1][fr_active]; (f != NULL); f = f->fr_next)
if (f->fr_ifa == ifp)
f->fr_ifa = (void *)-1;
#endif
RWLOCK_EXIT(&ipf_mutex);
fr_natsync(ifp);
}
#if defined(__sgi) && (IRIX < 60500)
static int no_output(ifp, m, s)
#else
# if TRU64 >= 1885
static int no_output (ifp, m, s, rt, cp)
char *cp;
# else
static int no_output(ifp, m, s, rt)
# endif
struct rtentry *rt;
#endif
struct ifnet *ifp;
struct mbuf *m;
#if defined(__NetBSD_Version__) && (__NetBSD_Version__ >= 499001100)
const struct sockaddr *s;
#else
struct sockaddr *s;
#endif
{
return 0;
}
#if defined(__sgi) && (IRIX < 60500)
static int write_output(ifp, m, s)
#else
# if TRU64 >= 1885
static int write_output (ifp, m, s, rt, cp)
char *cp;
# else
static int write_output(ifp, m, s, rt)
# endif
struct rtentry *rt;
#endif
struct ifnet *ifp;
struct mbuf *m;
#if defined(__NetBSD_Version__) && (__NetBSD_Version__ >= 499001100)
const struct sockaddr *s;
#else
struct sockaddr *s;
#endif
{
char fname[32];
mb_t *mb;
ip_t *ip;
int fd;
mb = (mb_t *)m;
ip = MTOD(mb, ip_t *);
#if (defined(NetBSD) && (NetBSD <= 1991011) && (NetBSD >= 199606)) || \
(defined(OpenBSD) && (OpenBSD >= 199603)) || defined(linux) || \
(defined(__FreeBSD__) && (__FreeBSD_version >= 501113))
sprintf(fname, "/tmp/%s", ifp->if_xname);
#else
sprintf(fname, "/tmp/%s%d", ifp->if_name, ifp->if_unit);
#endif
fd = open(fname, O_WRONLY|O_APPEND);
if (fd == -1) {
perror("open");
return -1;
}
write(fd, (char *)ip, ntohs(ip->ip_len));
close(fd);
return 0;
}
static void fr_setifpaddr(ifp, addr)
struct ifnet *ifp;
char *addr;
{
#ifdef __sgi
struct in_ifaddr *ifa;
#else
struct ifaddr *ifa;
#endif
#if defined(__NetBSD__) || defined(__OpenBSD__) || defined(__FreeBSD__)
if (ifp->if_addrlist.tqh_first != NULL)
#else
# ifdef __sgi
if (ifp->in_ifaddr != NULL)
# else
if (ifp->if_addrlist != NULL)
# endif
#endif
return;
ifa = (struct ifaddr *)malloc(sizeof(*ifa));
#if defined(__NetBSD__) || defined(__OpenBSD__) || defined(__FreeBSD__)
ifp->if_addrlist.tqh_first = ifa;
#else
# ifdef __sgi
ifp->in_ifaddr = ifa;
# else
ifp->if_addrlist = ifa;
# endif
#endif
if (ifa != NULL) {
struct sockaddr_in *sin;
#ifdef __sgi
sin = (struct sockaddr_in *)&ifa->ia_addr;
#else
sin = (struct sockaddr_in *)&ifa->ifa_addr;
#endif
sin->sin_addr.s_addr = inet_addr(addr);
if (sin->sin_addr.s_addr == 0)
abort();
}
}
struct ifnet *get_unit(name, v)
char *name;
int v;
{
struct ifnet *ifp, **ifpp, **old_ifneta;
char *addr;
#if (defined(NetBSD) && (NetBSD <= 1991011) && (NetBSD >= 199606)) || \
(defined(OpenBSD) && (OpenBSD >= 199603)) || defined(linux) || \
(defined(__FreeBSD__) && (__FreeBSD_version >= 501113))
if (name == NULL)
name = "anon0";
addr = strchr(name, '=');
if (addr != NULL)
*addr++ = '\0';
for (ifpp = ifneta; ifpp && (ifp = *ifpp); ifpp++) {
if (!strcmp(name, ifp->if_xname)) {
if (addr != NULL)
fr_setifpaddr(ifp, addr);
return ifp;
}
}
#else
char *s, ifname[LIFNAMSIZ+1];
if (name == NULL)
name = "anon0";
addr = strchr(name, '=');
if (addr != NULL)
*addr++ = '\0';
for (ifpp = ifneta; ifpp && (ifp = *ifpp); ifpp++) {
COPYIFNAME(v, ifp, ifname);
if (!strcmp(name, ifname)) {
if (addr != NULL)
fr_setifpaddr(ifp, addr);
return ifp;
}
}
#endif
if (!ifneta) {
ifneta = (struct ifnet **)malloc(sizeof(ifp) * 2);
if (!ifneta)
return NULL;
ifneta[1] = NULL;
ifneta[0] = (struct ifnet *)calloc(1, sizeof(*ifp));
if (!ifneta[0]) {
free(ifneta);
return NULL;
}
nifs = 1;
} else {
old_ifneta = ifneta;
nifs++;
ifneta = (struct ifnet **)realloc(ifneta,
(nifs + 1) * sizeof(ifp));
if (!ifneta) {
free(old_ifneta);
nifs = 0;
return NULL;
}
ifneta[nifs] = NULL;
ifneta[nifs - 1] = (struct ifnet *)malloc(sizeof(*ifp));
if (!ifneta[nifs - 1]) {
nifs--;
return NULL;
}
}
ifp = ifneta[nifs - 1];
#if defined(__NetBSD__) || defined(__OpenBSD__) || defined(__FreeBSD__)
TAILQ_INIT(&ifp->if_addrlist);
#endif
#if (defined(NetBSD) && (NetBSD <= 1991011) && (NetBSD >= 199606)) || \
(defined(OpenBSD) && (OpenBSD >= 199603)) || defined(linux) || \
(defined(__FreeBSD__) && (__FreeBSD_version >= 501113))
(void) strncpy(ifp->if_xname, name, sizeof(ifp->if_xname));
#else
for (s = name; *s && !ISDIGIT(*s); s++)
;
if (*s && ISDIGIT(*s)) {
ifp->if_unit = atoi(s);
ifp->if_name = (char *)malloc(s - name + 1);
(void) strncpy(ifp->if_name, name, s - name);
ifp->if_name[s - name] = '\0';
} else {
ifp->if_name = strdup(name);
ifp->if_unit = -1;
}
#endif
ifp->if_output = (void *)no_output;
if (addr != NULL) {
fr_setifpaddr(ifp, addr);
}
return ifp;
}
char *get_ifname(ifp)
struct ifnet *ifp;
{
static char ifname[LIFNAMSIZ];
#if defined(__OpenBSD__) || defined(__NetBSD__) || defined(linux) || \
(defined(__FreeBSD__) && (__FreeBSD_version >= 501113))
sprintf(ifname, "%s", ifp->if_xname);
#else
sprintf(ifname, "%s%d", ifp->if_name, ifp->if_unit);
#endif
return ifname;
}
void init_ifp()
{
struct ifnet *ifp, **ifpp;
char fname[32];
int fd;
#if (defined(NetBSD) && (NetBSD <= 1991011) && (NetBSD >= 199606)) || \
(defined(OpenBSD) && (OpenBSD >= 199603)) || defined(linux) || \
(defined(__FreeBSD__) && (__FreeBSD_version >= 501113))
for (ifpp = ifneta; ifpp && (ifp = *ifpp); ifpp++) {
ifp->if_output = (void *)write_output;
sprintf(fname, "/tmp/%s", ifp->if_xname);
fd = open(fname, O_WRONLY|O_CREAT|O_EXCL|O_TRUNC, 0600);
if (fd == -1)
perror("open");
else
close(fd);
}
#else
for (ifpp = ifneta; ifpp && (ifp = *ifpp); ifpp++) {
ifp->if_output = write_output;
sprintf(fname, "/tmp/%s%d", ifp->if_name, ifp->if_unit);
fd = open(fname, O_WRONLY|O_CREAT|O_EXCL|O_TRUNC, 0600);
if (fd == -1)
perror("open");
else
close(fd);
}
#endif
}
int fr_fastroute(m, mpp, fin, fdp)
mb_t *m, **mpp;
fr_info_t *fin;
frdest_t *fdp;
{
struct ifnet *ifp = fdp->fd_ifp;
ip_t *ip = fin->fin_ip;
int error = 0;
frentry_t *fr;
void *sifp;
if (!ifp)
return 0; /* no routing table out here */
fr = fin->fin_fr;
ip->ip_sum = 0;
if (fin->fin_out == 0) {
sifp = fin->fin_ifp;
fin->fin_ifp = ifp;
fin->fin_out = 1;
(void) fr_acctpkt(fin, NULL);
fin->fin_fr = NULL;
if (!fr || !(fr->fr_flags & FR_RETMASK)) {
u_32_t pass;
(void) fr_checkstate(fin, &pass);
}
switch (fr_checknatout(fin, NULL))
{
case 0 :
break;
case 1 :
ip->ip_sum = 0;
break;
case -1 :
error = -1;
goto done;
break;
}
fin->fin_ifp = sifp;
fin->fin_out = 0;
}
#if defined(__sgi) && (IRIX < 60500)
(*ifp->if_output)(ifp, (void *)ip, NULL);
# if TRU64 >= 1885
(*ifp->if_output)(ifp, (void *)m, NULL, 0, 0);
# else
(*ifp->if_output)(ifp, (void *)m, NULL, 0);
# endif
#endif
done:
return error;
}
int fr_send_reset(fin)
fr_info_t *fin;
{
verbose("- TCP RST sent\n");
return 0;
}
int fr_send_icmp_err(type, fin, dst)
int type;
fr_info_t *fin;
int dst;
{
verbose("- ICMP unreachable sent\n");
return 0;
}
void frsync(ifp)
void *ifp;
{
return;
}
void m_freem(m)
mb_t *m;
{
return;
}
void m_copydata(m, off, len, cp)
mb_t *m;
int off, len;
void *cp;
{
bcopy((char *)m + off, cp, len);
}
int ipfuiomove(buf, len, rwflag, uio)
void *buf;
int len, rwflag;
struct uio *uio;
{
int left, ioc, num, offset;
struct iovec *io;
char *start;
if (rwflag == UIO_READ) {
left = len;
ioc = 0;
offset = uio->uio_offset;
while ((left > 0) && (ioc < uio->uio_iovcnt)) {
io = uio->uio_iov + ioc;
num = io->iov_len;
if (num > left)
num = left;
start = (char *)io->iov_base + offset;
if (start > (char *)io->iov_base + io->iov_len) {
offset -= io->iov_len;
ioc++;
continue;
}
bcopy(buf, start, num);
uio->uio_resid -= num;
uio->uio_offset += num;
left -= num;
if (left > 0)
ioc++;
}
if (left > 0)
return EFAULT;
}
return 0;
}
u_32_t fr_newisn(fin)
fr_info_t *fin;
{
static int iss_seq_off = 0;
u_char hash[16];
u_32_t newiss;
MD5_CTX ctx;
/*
* Compute the base value of the ISS. It is a hash
* of (saddr, sport, daddr, dport, secret).
*/
MD5Init(&ctx);
MD5Update(&ctx, (u_char *) &fin->fin_fi.fi_src,
sizeof(fin->fin_fi.fi_src));
MD5Update(&ctx, (u_char *) &fin->fin_fi.fi_dst,
sizeof(fin->fin_fi.fi_dst));
MD5Update(&ctx, (u_char *) &fin->fin_dat, sizeof(fin->fin_dat));
/* MD5Update(&ctx, ipf_iss_secret, sizeof(ipf_iss_secret)); */
MD5Final(hash, &ctx);
memcpy(&newiss, hash, sizeof(newiss));
/*
* Now increment our "timer", and add it in to
* the computed value.
*
* XXX Use `addin'?
* XXX TCP_ISSINCR too large to use?
*/
iss_seq_off += 0x00010000;
newiss += iss_seq_off;
return newiss;
}
/* ------------------------------------------------------------------------ */
/* Function: fr_nextipid */
/* Returns: int - 0 == success, -1 == error (packet should be droppped) */
/* Parameters: fin(I) - pointer to packet information */
/* */
/* Returns the next IPv4 ID to use for this packet. */
/* ------------------------------------------------------------------------ */
EXTERN_INLINE u_short fr_nextipid(fin)
fr_info_t *fin;
{
static u_short ipid = 0;
u_short id;
MUTEX_ENTER(&ipf_rw);
id = ipid++;
MUTEX_EXIT(&ipf_rw);
return id;
}
EXTERN_INLINE void fr_checkv4sum(fin)
fr_info_t *fin;
{
if (fr_checkl4sum(fin) == -1)
fin->fin_flx |= FI_BAD;
}
#ifdef USE_INET6
EXTERN_INLINE void fr_checkv6sum(fin)
fr_info_t *fin;
{
if (fr_checkl4sum(fin) == -1)
fin->fin_flx |= FI_BAD;
}
#endif
/*
* See above for description, except that all addressing is in user space.
*/
int copyoutptr(src, dst, size)
void *src, *dst;
size_t size;
{
caddr_t ca;
bcopy(dst, (char *)&ca, sizeof(ca));
bcopy(src, ca, size);
return 0;
}
/*
* See above for description, except that all addressing is in user space.
*/
int copyinptr(src, dst, size)
void *src, *dst;
size_t size;
{
caddr_t ca;
bcopy(src, (char *)&ca, sizeof(ca));
bcopy(ca, dst, size);
return 0;
}
/*
* return the first IP Address associated with an interface
*/
int fr_ifpaddr(v, atype, ifptr, inp, inpmask)
int v, atype;
void *ifptr;
struct in_addr *inp, *inpmask;
{
struct ifnet *ifp = ifptr;
#ifdef __sgi
struct in_ifaddr *ifa;
#else
struct ifaddr *ifa;
#endif
#if defined(__NetBSD__) || defined(__OpenBSD__) || defined(__FreeBSD__)
ifa = ifp->if_addrlist.tqh_first;
#else
# ifdef __sgi
ifa = (struct in_ifaddr *)ifp->in_ifaddr;
# else
ifa = ifp->if_addrlist;
# endif
#endif
if (ifa != NULL) {
struct sockaddr_in *sin, mask;
mask.sin_addr.s_addr = 0xffffffff;
#ifdef __sgi
sin = (struct sockaddr_in *)&ifa->ia_addr;
#else
sin = (struct sockaddr_in *)&ifa->ifa_addr;
#endif
return fr_ifpfillv4addr(atype, sin, &mask, inp, inpmask);
}
return 0;
}
int ipfsync()
{
return 0;
}
/*
* This function is not meant to be random, rather just produce a
* sequence of numbers that isn't linear to show "randomness".
*/
u_32_t
ipf_random()
{
static int last = 0xa5a5a5a5;
static int calls = 0;
int number;
calls++;
/*
* These are deliberately chosen to ensure that there is some
* attempt to test whether the output covers the range in test n18.
*/
switch (calls)
{
case 1 :
number = 0;
break;
case 2 :
number = 4;
break;
case 3 :
number = 3999;
break;
case 4 :
number = 4000;
break;
case 5 :
number = 48999;
break;
case 6 :
number = 49000;
break;
default :
/*
* So why not use srand/rand/srandom/random? Because the
* actual values returned vary from platform to platform
* and what is needed is seomthing that is the same everywhere
* so that regression tests can work. Well, they could be
* built on each platform to suit but that's a whole lot of
* work for little gain given that we don't actually need
* random numbers here, just a spread to test the NAT code
* with.
*/
number = last;
last *= calls;
last++;
number ^= last;
break;
}
return number;
}