NetBSD/sys/netinet/ip_state.c

1665 lines
38 KiB
C
Raw Normal View History

2000-05-23 10:07:42 +04:00
/* $NetBSD: ip_state.c,v 1.26 2000/05/23 06:07:44 veego Exp $ */
/*
2000-05-03 15:12:03 +04:00
* Copyright (C) 1995-2000 by Darren Reed.
*
* Redistribution and use in source and binary forms are permitted
* provided that this notice is preserved and due credit is given
* to the original author and the contributors.
*/
#if !defined(lint)
#if defined(__NetBSD__)
2000-05-23 10:07:42 +04:00
static const char rcsid[] = "$NetBSD: ip_state.c,v 1.26 2000/05/23 06:07:44 veego Exp $";
#else
2000-05-03 15:12:03 +04:00
static const char sccsid[] = "@(#)ip_state.c 1.8 6/5/96 (C) 1993-2000 Darren Reed";
2000-05-23 10:07:42 +04:00
static const char rcsid[] = "@(#)Id: ip_state.c,v 2.30.2.9 2000/05/22 10:26:15 darrenr Exp";
#endif
#endif
1998-11-22 18:17:18 +03:00
#include <sys/errno.h>
#include <sys/types.h>
#include <sys/param.h>
#include <sys/file.h>
2000-05-03 15:12:03 +04:00
#if defined(__NetBSD__) && (NetBSD >= 199905) && !defined(IPFILTER_LKM) && \
defined(_KERNEL) && !defined(_LKM)
# include "opt_ipfilter_log.h"
#endif
#if defined(_KERNEL) && defined(__FreeBSD_version) && \
(__FreeBSD_version >= 400000) && !defined(KLD_MODULE)
#include "opt_inet6.h"
#endif
1997-11-14 15:40:06 +03:00
#if !defined(_KERNEL) && !defined(KERNEL) && !defined(__KERNEL__)
1998-11-22 18:17:18 +03:00
# include <stdio.h>
# include <stdlib.h>
# include <string.h>
1997-11-14 15:40:06 +03:00
#else
# ifdef linux
# include <linux/kernel.h>
# include <linux/module.h>
# endif
#endif
2000-05-03 15:12:03 +04:00
#if (defined(KERNEL) || defined(_KERNEL)) && (__FreeBSD_version >= 220000)
1997-05-25 16:40:11 +04:00
# include <sys/filio.h>
# include <sys/fcntl.h>
1999-12-12 14:11:15 +03:00
# if (__FreeBSD_version >= 300000) && !defined(IPFILTER_LKM)
# include "opt_ipfilter.h"
# endif
1997-05-25 16:40:11 +04:00
#else
# include <sys/ioctl.h>
#endif
1997-11-14 15:40:06 +03:00
#include <sys/time.h>
#include <sys/uio.h>
1997-11-14 15:40:06 +03:00
#ifndef linux
1999-12-12 14:11:15 +03:00
# include <sys/protosw.h>
1997-11-14 15:40:06 +03:00
#endif
#include <sys/socket.h>
2000-05-03 15:12:03 +04:00
#if (defined(_KERNEL) || defined(KERNEL)) && !defined(linux)
# include <sys/systm.h>
#endif
#if !defined(__SVR4) && !defined(__svr4__)
1997-11-14 15:40:06 +03:00
# ifndef linux
# include <sys/mbuf.h>
# endif
#else
1997-05-25 16:40:11 +04:00
# include <sys/filio.h>
# include <sys/byteorder.h>
1999-12-12 14:11:15 +03:00
# ifdef _KERNEL
# include <sys/dditypes.h>
# endif
# include <sys/stream.h>
# include <sys/kmem.h>
#endif
#include <net/if.h>
#ifdef sun
1999-12-12 14:11:15 +03:00
# include <net/af.h>
#endif
#include <net/route.h>
#include <netinet/in.h>
#include <netinet/in_systm.h>
#include <netinet/ip.h>
#include <netinet/tcp.h>
1997-11-14 15:40:06 +03:00
#ifndef linux
# include <netinet/ip_var.h>
# include <netinet/tcp_fsm.h>
#endif
#include <netinet/udp.h>
#include <netinet/ip_icmp.h>
1997-05-25 16:40:11 +04:00
#include "netinet/ip_compat.h"
1997-11-14 15:40:06 +03:00
#include <netinet/tcpip.h>
1997-05-25 16:40:11 +04:00
#include "netinet/ip_fil.h"
#include "netinet/ip_nat.h"
#include "netinet/ip_frag.h"
1997-07-05 09:38:14 +04:00
#include "netinet/ip_proxy.h"
1997-05-25 16:40:11 +04:00
#include "netinet/ip_state.h"
2000-05-03 15:12:03 +04:00
#ifdef USE_INET6
#include <netinet/icmp6.h>
#endif
1999-12-12 14:11:15 +03:00
#if (__FreeBSD_version >= 300000)
# include <sys/malloc.h>
# if (defined(_KERNEL) || defined(KERNEL)) && !defined(IPFILTER_LKM)
# include <sys/libkern.h>
# include <sys/systm.h>
# endif
#endif
#ifndef MIN
1999-12-12 14:11:15 +03:00
# define MIN(a,b) (((a)<(b))?(a):(b))
#endif
#define TCP_CLOSE (TH_FIN|TH_RST)
2000-05-03 15:12:03 +04:00
static ipstate_t **ips_table = NULL;
static ipstate_t *ips_list = NULL;
static int ips_num = 0;
static ips_stat_t ips_stats;
#if (SOLARIS || defined(__sgi)) && defined(_KERNEL)
1999-12-12 14:11:15 +03:00
extern KRWLOCK_T ipf_state, ipf_mutex;
1998-07-12 19:23:59 +04:00
extern kmutex_t ipf_rw;
#endif
2000-05-03 15:12:03 +04:00
#ifdef USE_INET6
static frentry_t *fr_checkicmp6matchingstate __P((ip6_t *, fr_info_t *));
#endif
static int fr_matchsrcdst __P((ipstate_t *, union i6addr, union i6addr,
1999-12-12 14:11:15 +03:00
fr_info_t *, tcphdr_t *));
static frentry_t *fr_checkicmpmatchingstate __P((ip_t *, fr_info_t *));
2000-05-03 15:12:03 +04:00
static int fr_matchicmpqueryreply __P((int, ipstate_t *, icmphdr_t *));
static int fr_state_flush __P((int));
static ips_stat_t *fr_statetstats __P((void));
1999-12-12 14:11:15 +03:00
static void fr_delstate __P((ipstate_t *));
2000-05-03 15:12:03 +04:00
static int fr_state_remove __P((caddr_t));
int fr_stputent __P((caddr_t));
int fr_stgetent __P((caddr_t));
void fr_stinsert __P((ipstate_t *));
#define FIVE_DAYS (2 * 5 * 86400) /* 5 days: half closed session */
1999-12-12 14:11:15 +03:00
#define TCP_MSL 240 /* 2 minutes */
u_long fr_tcpidletimeout = FIVE_DAYS,
1999-12-12 14:11:15 +03:00
fr_tcpclosewait = 2 * TCP_MSL,
fr_tcplastack = 2 * TCP_MSL,
fr_tcptimeout = 2 * TCP_MSL,
fr_tcpclosed = 1,
1999-12-12 14:11:15 +03:00
fr_udptimeout = 240,
fr_icmptimeout = 120;
1999-12-12 14:11:15 +03:00
int fr_statemax = IPSTATE_MAX,
fr_statesize = IPSTATE_SIZE;
2000-05-03 15:12:03 +04:00
int fr_state_doflush = 0,
fr_state_lock = 0;
1999-12-12 14:11:15 +03:00
2000-05-03 15:12:03 +04:00
static int icmpreplytype4[ICMP_MAXTYPE + 1];
1999-12-12 14:11:15 +03:00
int fr_stateinit()
{
2000-05-03 15:12:03 +04:00
int i;
1999-12-12 14:11:15 +03:00
KMALLOCS(ips_table, ipstate_t **, fr_statesize * sizeof(ipstate_t *));
if (ips_table != NULL)
bzero((char *)ips_table, fr_statesize * sizeof(ipstate_t *));
else
return -1;
2000-05-03 15:12:03 +04:00
/* fill icmp reply type table */
for (i = 0; i <= ICMP_MAXTYPE; i++)
icmpreplytype4[i] = -1;
icmpreplytype4[ICMP_ECHO] = ICMP_ECHOREPLY;
icmpreplytype4[ICMP_TSTAMP] = ICMP_TSTAMPREPLY;
icmpreplytype4[ICMP_IREQ] = ICMP_IREQREPLY;
icmpreplytype4[ICMP_MASKREQ] = ICMP_MASKREPLY;
1999-12-12 14:11:15 +03:00
return 0;
}
static ips_stat_t *fr_statetstats()
{
ips_stats.iss_active = ips_num;
ips_stats.iss_table = ips_table;
2000-05-03 15:12:03 +04:00
ips_stats.iss_list = ips_list;
return &ips_stats;
}
1997-11-14 15:40:06 +03:00
/*
* flush state tables. two actions currently defined:
* which == 0 : flush all state table entries
* which == 1 : flush TCP connections which have started to close but are
* stuck for some reason.
*/
static int fr_state_flush(which)
1997-11-14 15:40:06 +03:00
int which;
{
2000-05-03 15:12:03 +04:00
register ipstate_t *is, **isp;
1997-11-14 15:40:06 +03:00
#if defined(_KERNEL) && !SOLARIS
int s;
#endif
int delete, removed = 0;
SPL_NET(s);
2000-05-03 15:12:03 +04:00
for (isp = &ips_list; (is = *isp); ) {
delete = 0;
1997-11-14 15:40:06 +03:00
2000-05-03 15:12:03 +04:00
switch (which)
{
case 0 :
delete = 1;
break;
case 1 :
if (is->is_p != IPPROTO_TCP)
1997-11-14 15:40:06 +03:00
break;
2000-05-03 15:12:03 +04:00
if ((is->is_state[0] != TCPS_ESTABLISHED) ||
(is->is_state[1] != TCPS_ESTABLISHED))
delete = 1;
break;
}
1997-11-14 15:40:06 +03:00
2000-05-03 15:12:03 +04:00
if (delete) {
if (is->is_p == IPPROTO_TCP)
ips_stats.iss_fin++;
else
ips_stats.iss_expire++;
1997-11-14 15:40:06 +03:00
#ifdef IPFILTER_LOG
2000-05-03 15:12:03 +04:00
ipstate_log(is, ISL_FLUSH);
1997-11-14 15:40:06 +03:00
#endif
2000-05-03 15:12:03 +04:00
fr_delstate(is);
removed++;
} else
isp = &is->is_next;
}
1997-11-14 15:40:06 +03:00
SPL_X(s);
return removed;
}
2000-05-03 15:12:03 +04:00
static int fr_state_remove(data)
caddr_t data;
{
ipstate_t *sp, st;
int error;
sp = &st;
error = IRCOPYPTR(data, (caddr_t)&st, sizeof(st));
if (error)
return EFAULT;
for (sp = ips_list; sp; sp = sp->is_next)
if ((sp->is_p == st.is_p) && (sp->is_v == st.is_v) &&
!bcmp(&sp->is_src, &st.is_src, sizeof(st.is_src)) &&
!bcmp(&sp->is_dst, &st.is_src, sizeof(st.is_dst)) &&
!bcmp(&sp->is_ps, &st.is_ps, sizeof(st.is_ps))) {
WRITE_ENTER(&ipf_state);
#ifdef IPFILTER_LOG
ipstate_log(sp, ISL_REMOVE);
#endif
fr_delstate(sp);
RWLOCK_EXIT(&ipf_state);
return 0;
}
return ESRCH;
}
1997-05-25 16:40:11 +04:00
int fr_state_ioctl(data, cmd, mode)
caddr_t data;
#if defined(__NetBSD__) || defined(__OpenBSD__)
1997-07-05 09:38:14 +04:00
u_long cmd;
#else
1997-05-25 16:40:11 +04:00
int cmd;
1997-07-05 09:38:14 +04:00
#endif
1997-05-25 16:40:11 +04:00
int mode;
{
2000-05-03 15:12:03 +04:00
int arg, ret, error = 0;
1997-11-14 15:40:06 +03:00
1997-05-25 16:40:11 +04:00
switch (cmd)
{
2000-05-03 15:12:03 +04:00
case SIOCDELST :
error = fr_state_remove(data);
break;
1997-11-14 15:40:06 +03:00
case SIOCIPFFL :
2000-05-03 15:12:03 +04:00
error = IRCOPY(data, (caddr_t)&arg, sizeof(arg));
if (error)
break;
1997-11-14 15:40:06 +03:00
if (arg == 0 || arg == 1) {
2000-05-03 15:12:03 +04:00
WRITE_ENTER(&ipf_state);
1997-11-14 15:40:06 +03:00
ret = fr_state_flush(arg);
2000-05-03 15:12:03 +04:00
RWLOCK_EXIT(&ipf_state);
error = IWCOPY((caddr_t)&ret, data, sizeof(ret));
1997-11-14 15:40:06 +03:00
} else
error = EINVAL;
break;
1999-12-28 10:14:53 +03:00
#ifdef IPFILTER_LOG
case SIOCIPFFB :
if (!(mode & FWRITE))
error = EPERM;
2000-05-21 22:45:53 +04:00
else {
int tmp;
tmp = ipflog_clear(IPL_LOGSTATE);
IWCOPY((char *)&tmp, data, sizeof(tmp));
}
1999-12-28 10:14:53 +03:00
break;
#endif
2000-05-03 15:12:03 +04:00
case SIOCGETFS :
error = IWCOPYPTR((caddr_t)fr_statetstats(), data,
sizeof(ips_stat_t));
1997-05-25 16:40:11 +04:00
break;
case FIONREAD :
#ifdef IPFILTER_LOG
2000-05-03 15:12:03 +04:00
error = IWCOPY((caddr_t)&iplused[IPL_LOGSTATE], (caddr_t)data,
sizeof(iplused[IPL_LOGSTATE]));
1997-05-25 16:40:11 +04:00
#endif
break;
2000-05-03 15:12:03 +04:00
case SIOCSTLCK :
error = fr_lock(data, &fr_state_lock);
break;
case SIOCSTPUT :
if (!fr_state_lock) {
error = EACCES;
break;
}
error = fr_stputent(data);
break;
case SIOCSTGET :
if (!fr_state_lock) {
error = EACCES;
break;
}
error = fr_stgetent(data);
break;
1997-05-25 16:40:11 +04:00
default :
1999-12-12 14:11:15 +03:00
error = EINVAL;
break;
1997-05-25 16:40:11 +04:00
}
1997-11-14 15:40:06 +03:00
return error;
1997-05-25 16:40:11 +04:00
}
2000-05-03 15:12:03 +04:00
int fr_stgetent(data)
caddr_t data;
{
register ipstate_t *is, *isn;
ipstate_save_t ips, *ipsp;
int error;
error = IRCOPY(data, (caddr_t)&ipsp, sizeof(ipsp));
if (error)
return EFAULT;
error = IRCOPY((caddr_t)ipsp, (caddr_t)&ips, sizeof(ips));
if (error)
return EFAULT;
isn = ips.ips_next;
if (!isn) {
isn = ips_list;
if (isn == NULL) {
if (ips.ips_next == NULL)
return ENOENT;
return 0;
}
} else {
/*
* Make sure the pointer we're copying from exists in the
* current list of entries. Security precaution to prevent
* copying of random kernel data.
*/
for (is = ips_list; is; is = is->is_next)
if (is == isn)
break;
if (!is)
return ESRCH;
}
ips.ips_next = isn->is_next;
bcopy((char *)isn, (char *)&ips.ips_is, sizeof(ips.ips_is));
if (isn->is_rule)
bcopy((char *)isn->is_rule, (char *)&ips.ips_fr,
sizeof(ips.ips_fr));
error = IWCOPY((caddr_t)&ips, ipsp, sizeof(ips));
if (error)
return EFAULT;
return 0;
}
int fr_stputent(data)
caddr_t data;
{
register ipstate_t *is, *isn;
ipstate_save_t ips, *ipsp;
frentry_t *fr;
int error;
error = IRCOPY(data, (caddr_t)&ipsp, sizeof(ipsp));
if (error)
return EFAULT;
error = IRCOPY((caddr_t)ipsp, (caddr_t)&ips, sizeof(ips));
if (error)
return EFAULT;
KMALLOC(isn, ipstate_t *);
if (isn == NULL)
return ENOMEM;
bcopy((char *)&ips.ips_is, (char *)isn, sizeof(*isn));
fr = isn->is_rule;
if (fr != NULL) {
if (isn->is_flags & FI_NEWFR) {
KMALLOC(fr, frentry_t *);
if (fr == NULL) {
KFREE(isn);
return ENOMEM;
}
bcopy((char *)&ips.ips_fr, (char *)fr, sizeof(*fr));
isn->is_rule = fr;
ips.ips_is.is_rule = fr;
error = IWCOPY((caddr_t)&ips, ipsp, sizeof(ips));
if (error) {
KFREE(isn);
KFREE(fr);
return EFAULT;
}
} else {
for (is = ips_list; is; is = is->is_next)
if (is->is_rule == fr)
break;
if (!is) {
KFREE(isn);
return ESRCH;
}
}
}
fr_stinsert(isn);
return 0;
}
void fr_stinsert(is)
register ipstate_t *is;
{
register u_int hv = is->is_hv;
MUTEX_INIT(&is->is_lock, "ipf state entry", NULL);
is->is_ifname[0][sizeof(is->is_ifname[0]) - 1] = '\0';
if (is->is_ifname[0][0] != '\0') {
is->is_ifp[0] = GETUNIT(is->is_ifname[0], is->is_v);
}
is->is_ifname[1][sizeof(is->is_ifname[0]) - 1] = '\0';
if (is->is_ifname[1][0] != '\0') {
is->is_ifp[1] = GETUNIT(is->is_ifname[1], is->is_v);
}
/*
* add into list table.
*/
if (ips_list)
ips_list->is_pnext = &is->is_next;
is->is_pnext = &ips_list;
is->is_next = ips_list;
ips_list = is;
if (ips_table[hv])
ips_table[hv]->is_phnext = &is->is_hnext;
else
ips_stats.iss_inuse++;
is->is_phnext = ips_table + hv;
is->is_hnext = ips_table[hv];
ips_table[hv] = is;
}
/*
* Create a new ipstate structure and hang it off the hash table.
*/
1999-12-12 14:11:15 +03:00
ipstate_t *fr_addstate(ip, fin, flags)
ip_t *ip;
fr_info_t *fin;
1999-12-12 14:11:15 +03:00
u_int flags;
{
2000-05-03 15:12:03 +04:00
register tcphdr_t *tcp = NULL;
register ipstate_t *is;
register u_int hv;
1999-12-12 14:11:15 +03:00
ipstate_t ips;
u_int pass;
2000-05-03 15:12:03 +04:00
int out;
2000-05-03 15:12:03 +04:00
if (fr_state_lock || (fin->fin_off & IP_OFFMASK) ||
(fin->fin_fi.fi_fl & FI_SHORT))
1999-12-12 14:11:15 +03:00
return NULL;
if (ips_num == fr_statemax) {
ips_stats.iss_max++;
1999-12-12 14:11:15 +03:00
fr_state_doflush = 1;
return NULL;
}
2000-05-03 15:12:03 +04:00
out = fin->fin_out;
1999-12-12 14:11:15 +03:00
is = &ips;
bzero((char *)is, sizeof(*is));
ips.is_age = 1;
ips.is_state[0] = 0;
ips.is_state[1] = 0;
/*
* Copy and calculate...
*/
2000-05-03 15:12:03 +04:00
hv = (is->is_p = fin->fin_fi.fi_p);
is->is_src = fin->fin_fi.fi_src;
hv += is->is_saddr;
is->is_dst = fin->fin_fi.fi_dst;
hv += is->is_daddr;
#ifdef USE_INET6
if (fin->fin_v == 6) {
if (is->is_p == IPPROTO_ICMPV6) {
if (IN6_IS_ADDR_MULTICAST(&is->is_dst.in6))
flags |= FI_W_DADDR;
if (out)
hv -= is->is_daddr;
else
hv -= is->is_saddr;
}
}
#endif
2000-05-03 15:12:03 +04:00
switch (is->is_p)
{
2000-05-03 15:12:03 +04:00
#ifdef USE_INET6
case IPPROTO_ICMPV6 :
#endif
case IPPROTO_ICMP :
{
struct icmp *ic = (struct icmp *)fin->fin_dp;
2000-05-03 15:12:03 +04:00
#ifdef USE_INET6
if ((is->is_p == IPPROTO_ICMPV6) &&
((ic->icmp_type & ICMP6_INFOMSG_MASK) == 0))
return NULL;
#endif
switch (ic->icmp_type)
{
2000-05-03 15:12:03 +04:00
#ifdef USE_INET6
case ICMP6_ECHO_REQUEST :
is->is_icmp.ics_type = ICMP6_ECHO_REPLY;
hv += (is->is_icmp.ics_id = ic->icmp_id);
hv += (is->is_icmp.ics_seq = ic->icmp_seq);
break;
2000-05-03 15:12:03 +04:00
case ICMP6_MEMBERSHIP_QUERY :
case ND_ROUTER_SOLICIT :
case ND_NEIGHBOR_SOLICIT :
is->is_icmp.ics_type = ic->icmp_type + 1;
break;
break;
#endif
case ICMP_ECHO :
case ICMP_TSTAMP :
case ICMP_IREQ :
case ICMP_MASKREQ :
2000-05-03 15:12:03 +04:00
is->is_icmp.ics_type = ic->icmp_type;
hv += (is->is_icmp.ics_id = ic->icmp_id);
hv += (is->is_icmp.ics_seq = ic->icmp_seq);
break;
default :
1999-12-12 14:11:15 +03:00
return NULL;
}
2000-05-03 15:12:03 +04:00
ATOMIC_INCL(ips_stats.iss_icmp);
is->is_age = fr_icmptimeout;
break;
}
case IPPROTO_TCP :
{
2000-05-03 15:12:03 +04:00
tcp = (tcphdr_t *)fin->fin_dp;
2000-05-23 10:07:42 +04:00
if (tcp->th_flags & TH_RST)
return NULL;
/*
* The endian of the ports doesn't matter, but the ack and
* sequence numbers do as we do mathematics on them later.
*/
1999-12-12 14:11:15 +03:00
is->is_dport = tcp->th_dport;
is->is_sport = tcp->th_sport;
if ((flags & (FI_W_DPORT|FI_W_SPORT)) == 0) {
hv += tcp->th_dport;
hv += tcp->th_sport;
}
2000-05-03 15:12:03 +04:00
is->is_send = ntohl(tcp->th_seq) + ip->ip_len -
fin->fin_hlen - (tcp->th_off << 2) +
((tcp->th_flags & TH_SYN) ? 1 : 0) +
((tcp->th_flags & TH_FIN) ? 1 : 0);
is->is_maxsend = is->is_send;
1999-12-12 14:11:15 +03:00
is->is_dend = 0;
2000-05-03 15:12:03 +04:00
is->is_maxdwin = 1;
1999-12-12 14:11:15 +03:00
is->is_maxswin = ntohs(tcp->th_win);
if (is->is_maxswin == 0)
is->is_maxswin = 1;
/*
1997-11-14 15:40:06 +03:00
* If we're creating state for a starting connection, start the
* timer on it as we'll never see an error if it fails to
* connect.
*/
2000-05-03 15:12:03 +04:00
ATOMIC_INCL(ips_stats.iss_tcp);
break;
}
case IPPROTO_UDP :
{
2000-05-03 15:12:03 +04:00
tcp = (tcphdr_t *)fin->fin_dp;
1999-12-12 14:11:15 +03:00
is->is_dport = tcp->th_dport;
is->is_sport = tcp->th_sport;
if ((flags & (FI_W_DPORT|FI_W_SPORT)) == 0) {
hv += tcp->th_dport;
hv += tcp->th_sport;
}
2000-05-03 15:12:03 +04:00
ATOMIC_INCL(ips_stats.iss_udp);
is->is_age = fr_udptimeout;
break;
}
default :
1999-12-12 14:11:15 +03:00
return NULL;
}
1999-12-12 14:11:15 +03:00
KMALLOC(is, ipstate_t *);
if (is == NULL) {
2000-05-03 15:12:03 +04:00
ATOMIC_INCL(ips_stats.iss_nomem);
1999-12-12 14:11:15 +03:00
return NULL;
}
bcopy((char *)&ips, (char *)is, sizeof(*is));
1999-12-12 14:11:15 +03:00
hv %= fr_statesize;
2000-05-03 15:12:03 +04:00
is->is_hv = hv;
1999-12-12 14:11:15 +03:00
is->is_rule = fin->fin_fr;
if (is->is_rule != NULL) {
2000-05-03 15:12:03 +04:00
ATOMIC_INC32(is->is_rule->fr_ref);
1999-12-12 14:11:15 +03:00
pass = is->is_rule->fr_flags;
} else
pass = fr_flags;
1998-07-12 19:23:59 +04:00
WRITE_ENTER(&ipf_state);
is->is_pass = pass;
is->is_pkts = 1;
2000-05-03 15:12:03 +04:00
is->is_bytes = fin->fin_dlen + fin->fin_hlen;
/*
1998-07-12 19:23:59 +04:00
* We want to check everything that is a property of this packet,
* but we don't (automatically) care about it's fragment status as
* this may change.
*/
2000-05-03 15:12:03 +04:00
is->is_v = fin->fin_fi.fi_v;
1998-07-12 19:23:59 +04:00
is->is_opt = fin->fin_fi.fi_optmsk;
is->is_optmsk = 0xffffffff;
is->is_sec = fin->fin_fi.fi_secmsk;
is->is_secmsk = 0xffff;
is->is_auth = fin->fin_fi.fi_auth;
is->is_authmsk = 0xffff;
1999-12-12 14:11:15 +03:00
is->is_flags = fin->fin_fi.fi_fl & FI_CMP;
is->is_flags |= FI_CMP << 4;
2000-05-03 15:12:03 +04:00
is->is_flags |= flags & (FI_WILDP|FI_WILDA);
is->is_ifp[1 - out] = NULL;
is->is_ifp[out] = fin->fin_ifp;
#ifdef _KERNEL
strncpy(is->is_ifname[out], IFNAME(fin->fin_ifp), IFNAMSIZ);
#endif
is->is_ifname[1 - out][0] = '\0';
if (pass & FR_LOGFIRST)
is->is_pass &= ~(FR_LOGFIRST|FR_LOG);
2000-05-03 15:12:03 +04:00
fr_stinsert(is);
ips_num++;
if (is->is_p == IPPROTO_TCP) {
MUTEX_ENTER(&is->is_lock);
fr_tcp_age(&is->is_age, is->is_state, fin,
tcp->th_sport == is->is_sport);
MUTEX_EXIT(&is->is_lock);
}
1997-05-25 16:40:11 +04:00
#ifdef IPFILTER_LOG
ipstate_log(is, ISL_NEW);
#endif
1998-07-12 19:23:59 +04:00
RWLOCK_EXIT(&ipf_state);
2000-05-03 15:12:03 +04:00
fin->fin_rev = IP6NEQ(is->is_dst, fin->fin_fi.fi_dst);
1997-05-25 16:40:11 +04:00
if (fin->fin_fi.fi_fl & FI_FRAG)
ipfr_newfrag(ip, fin, pass ^ FR_KEEPSTATE);
1999-12-12 14:11:15 +03:00
return is;
}
1999-12-12 14:11:15 +03:00
/*
* check to see if a packet with TCP headers fits within the TCP window.
* change timeout depending on whether new packet is a SYN-ACK returning for a
* SYN or a RST or FIN which indicate time to close up shop.
*/
int fr_tcpstate(is, fin, ip, tcp)
2000-03-30 16:51:13 +04:00
ipstate_t *is;
fr_info_t *fin;
ip_t *ip;
tcphdr_t *tcp;
{
2000-03-30 16:51:13 +04:00
tcp_seq seq, ack, end;
int ackskew;
1999-12-12 14:11:15 +03:00
tcpdata_t *fdata, *tdata;
u_short win, maxwin;
int ret = 0;
int source;
/*
* Find difference between last checked packet and this packet.
*/
2000-05-03 15:12:03 +04:00
source = IP6EQ(fin->fin_fi.fi_src, is->is_src);
1999-12-12 14:11:15 +03:00
fdata = &is->is_tcp.ts_data[!source];
tdata = &is->is_tcp.ts_data[source];
seq = ntohl(tcp->th_seq);
ack = ntohl(tcp->th_ack);
1999-12-12 14:11:15 +03:00
win = ntohs(tcp->th_win);
2000-05-03 15:12:03 +04:00
end = seq + fin->fin_dlen - (tcp->th_off << 2) +
1999-12-12 14:11:15 +03:00
((tcp->th_flags & TH_SYN) ? 1 : 0) +
2000-02-02 00:29:15 +03:00
((tcp->th_flags & TH_FIN) ? 1 : 0);
1997-07-05 09:38:14 +04:00
1999-12-12 14:11:15 +03:00
if (fdata->td_end == 0) {
/*
* Must be a (outgoing) SYN-ACK in reply to a SYN.
*/
fdata->td_end = end;
fdata->td_maxwin = 1;
fdata->td_maxend = end + 1;
}
1999-12-12 14:11:15 +03:00
if (!(tcp->th_flags & TH_ACK)) { /* Pretend an ack was sent */
ack = tdata->td_end;
} else if (((tcp->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) &&
(ack == 0)) {
/* gross hack to get around certain broken tcp stacks */
ack = tdata->td_end;
}
1999-12-12 14:11:15 +03:00
if (seq == end)
seq = end = fdata->td_end;
maxwin = tdata->td_maxwin;
ackskew = tdata->td_end - ack;
#define SEQ_GE(a,b) ((int)((a) - (b)) >= 0)
#define SEQ_GT(a,b) ((int)((a) - (b)) > 0)
if ((SEQ_GE(fdata->td_maxend, end)) &&
2000-02-02 00:29:15 +03:00
(SEQ_GE(seq, fdata->td_end - maxwin)) &&
1999-12-12 14:11:15 +03:00
/* XXX what about big packets */
#define MAXACKWINDOW 66000
(ackskew >= -MAXACKWINDOW) &&
(ackskew <= MAXACKWINDOW)) {
/* if ackskew < 0 then this should be due to fragented
* packets. There is no way to know the length of the
* total packet in advance.
* We do know the total length from the fragment cache though.
* Note however that there might be more sessions with
* exactly the same source and destination paramters in the
* state cache (and source and destination is the only stuff
* that is saved in the fragment cache). Note further that
* some TCP connections in the state cache are hashed with
* sport and dport as well which makes it not worthwhile to
* look for them.
* Thus, when ackskew is negative but still seems to belong
* to this session, we bump up the destinations end value.
*/
if (ackskew < 0)
tdata->td_end = ack;
/* update max window seen */
if (fdata->td_maxwin < win)
fdata->td_maxwin = win;
if (SEQ_GT(end, fdata->td_end))
fdata->td_end = end;
if (SEQ_GE(ack + win, tdata->td_maxend)) {
tdata->td_maxend = ack + win;
if (win == 0)
tdata->td_maxend++;
}
1999-12-12 14:11:15 +03:00
2000-05-03 15:12:03 +04:00
ATOMIC_INCL(ips_stats.iss_hits);
1997-05-25 16:40:11 +04:00
is->is_pkts++;
2000-05-03 15:12:03 +04:00
is->is_bytes += fin->fin_dlen + fin->fin_hlen;
/*
* Nearing end of connection, start timeout.
*/
2000-05-03 15:12:03 +04:00
MUTEX_ENTER(&is->is_lock);
fr_tcp_age(&is->is_age, is->is_state, fin, source);
MUTEX_EXIT(&is->is_lock);
1999-12-12 14:11:15 +03:00
ret = 1;
}
1999-12-12 14:11:15 +03:00
return ret;
}
1999-12-12 14:11:15 +03:00
static int fr_matchsrcdst(is, src, dst, fin, tcp)
ipstate_t *is;
2000-05-03 15:12:03 +04:00
union i6addr src, dst;
fr_info_t *fin;
1999-12-12 14:11:15 +03:00
tcphdr_t *tcp;
{
1999-12-12 14:11:15 +03:00
int ret = 0, rev, out, flags;
u_short sp, dp;
void *ifp;
2000-05-03 15:12:03 +04:00
rev = fin->fin_rev = IP6NEQ(is->is_dst, dst);
ifp = fin->fin_ifp;
out = fin->fin_out;
1999-12-12 14:11:15 +03:00
if (tcp != NULL) {
flags = is->is_flags;
sp = tcp->th_sport;
dp = tcp->th_dport;
} else {
2000-05-03 15:12:03 +04:00
flags = is->is_flags & FI_WILDA;
1999-12-12 14:11:15 +03:00
sp = 0;
dp = 0;
}
if (rev == 0) {
if (!out) {
2000-05-03 15:12:03 +04:00
if (is->is_ifpin == NULL || is->is_ifpin == ifp)
1999-12-12 14:11:15 +03:00
ret = 1;
} else {
1999-12-12 14:11:15 +03:00
if (is->is_ifpout == NULL || is->is_ifpout == ifp)
ret = 1;
}
} else {
if (out) {
2000-05-03 15:12:03 +04:00
if (is->is_ifpin == NULL || is->is_ifpin == ifp)
1999-12-12 14:11:15 +03:00
ret = 1;
} else {
1999-12-12 14:11:15 +03:00
if (is->is_ifpout == NULL || is->is_ifpout == ifp)
ret = 1;
}
}
1999-12-12 14:11:15 +03:00
if (ret == 0)
return 0;
ret = 0;
1999-12-12 14:11:15 +03:00
if (rev == 0) {
2000-05-03 15:12:03 +04:00
if (
(IP6EQ(is->is_dst, dst) || (flags & FI_W_DADDR)) &&
(IP6EQ(is->is_src, src) || (flags & FI_W_SADDR)) &&
1999-12-12 14:11:15 +03:00
(!tcp || ((sp == is->is_sport || flags & FI_W_SPORT) &&
(dp == is->is_dport || flags & FI_W_DPORT)))) {
ret = 1;
}
} else {
2000-05-03 15:12:03 +04:00
if (
(IP6EQ(is->is_dst, src) || (flags & FI_W_DADDR)) &&
(IP6EQ(is->is_src, dst) || (flags & FI_W_SADDR)) &&
1999-12-12 14:11:15 +03:00
(!tcp || ((sp == is->is_dport || flags & FI_W_DPORT) &&
(dp == is->is_sport || flags & FI_W_SPORT)))) {
ret = 1;
}
}
1999-12-12 14:11:15 +03:00
if (ret == 0)
return 0;
/*
* Whether or not this should be here, is questionable, but the aim
* is to get this out of the main line.
*/
1999-12-12 14:11:15 +03:00
if (tcp == NULL)
flags = is->is_flags & (FI_CMP|(FI_CMP<<4));
if (((fin->fin_fi.fi_fl & (flags >> 4)) != (flags & FI_CMP)) ||
((fin->fin_fi.fi_optmsk & is->is_optmsk) != is->is_opt) ||
((fin->fin_fi.fi_secmsk & is->is_secmsk) != is->is_sec) ||
((fin->fin_fi.fi_auth & is->is_authmsk) != is->is_auth))
return 0;
if ((flags & (FI_W_SPORT|FI_W_DPORT))) {
if ((flags & FI_W_SPORT) != 0) {
if (rev == 0) {
is->is_sport = sp;
is->is_send = htonl(tcp->th_seq);
} else {
is->is_sport = dp;
is->is_send = htonl(tcp->th_ack);
}
is->is_maxsend = is->is_send + 1;
} else if ((flags & FI_W_DPORT) != 0) {
if (rev == 0) {
is->is_dport = dp;
is->is_dend = htonl(tcp->th_ack);
} else {
is->is_dport = sp;
is->is_dend = htonl(tcp->th_seq);
}
is->is_maxdend = is->is_dend + 1;
}
is->is_flags &= ~(FI_W_SPORT|FI_W_DPORT);
}
1999-12-12 14:11:15 +03:00
2000-05-03 15:12:03 +04:00
ret = -1;
1999-12-12 14:11:15 +03:00
if (!rev) {
2000-05-03 15:12:03 +04:00
if (out) {
1999-12-12 14:11:15 +03:00
if (!is->is_ifpout)
2000-05-03 15:12:03 +04:00
ret = 1;
1999-12-12 14:11:15 +03:00
} else {
if (!is->is_ifpin)
2000-05-03 15:12:03 +04:00
ret = 0;
1999-12-12 14:11:15 +03:00
}
} else {
2000-05-03 15:12:03 +04:00
if (out) {
1999-12-12 14:11:15 +03:00
if (!is->is_ifpin)
2000-05-03 15:12:03 +04:00
ret = 0;
1999-12-12 14:11:15 +03:00
} else {
if (!is->is_ifpout)
2000-05-03 15:12:03 +04:00
ret = 1;
1999-12-12 14:11:15 +03:00
}
}
2000-05-03 15:12:03 +04:00
if (ret >= 0) {
is->is_ifp[ret] = ifp;
#ifdef _KERNEL
strncpy(is->is_ifname[out], IFNAME(fin->fin_ifp),
sizeof(is->is_ifname[1]));
#endif
}
#ifdef _KERNEL
if (ret >= 0) {
strncpy(is->is_ifname[out], IFNAME(fin->fin_ifp),
sizeof(is->is_ifname[1]));
}
#endif
1999-12-12 14:11:15 +03:00
return 1;
}
2000-05-03 15:12:03 +04:00
static int fr_matchicmpqueryreply(v, is, icmp)
int v;
ipstate_t *is;
icmphdr_t *icmp;
{
if (v == 4) {
2000-05-21 22:45:53 +04:00
/*
* If we matched its type on the way in, then when going out
* it will still be the same type.
*/
if (((icmp->icmp_type == is->is_type) ||
(icmpreplytype4[is->is_type] == icmp->icmp_type)) &&
2000-05-03 15:12:03 +04:00
(icmp->icmp_id == is->is_icmp.ics_id) &&
(icmp->icmp_seq == is->is_icmp.ics_seq)) {
return 1;
};
}
#ifdef USE_INET6
else if (is->is_v == 6) {
if ((is->is_type == ICMP6_ECHO_REPLY) &&
(icmp->icmp_type == ICMP6_ECHO_REQUEST) &&
(icmp->icmp_id == is->is_icmp.ics_id) &&
(icmp->icmp_seq == is->is_icmp.ics_seq)) {
return 1;
};
}
#endif
return 0;
}
2000-05-21 22:45:53 +04:00
static frentry_t *fr_checkicmpmatchingstate(ip, fin)
1999-12-12 14:11:15 +03:00
ip_t *ip;
fr_info_t *fin;
{
2000-05-03 15:12:03 +04:00
register ipstate_t *is, **isp;
register u_short sport, dport;
register u_char pr;
union i6addr dst, src;
1999-12-12 14:11:15 +03:00
struct icmp *ic;
u_short savelen;
fr_info_t ofin;
tcphdr_t *tcp;
icmphdr_t *icmp;
frentry_t *fr;
ip_t *oip;
int type;
1999-12-28 10:14:53 +03:00
u_int hv;
1999-12-12 14:11:15 +03:00
2000-02-02 00:29:15 +03:00
/*
* Does it at least have the return (basic) IP header ?
1999-12-12 14:11:15 +03:00
* Only a basic IP header (no options) should be with
* an ICMP error header.
*/
2000-05-03 15:12:03 +04:00
if (((ip->ip_v != 4) && (ip->ip_hl != 5)) ||
(fin->fin_plen < ICMPERR_MINPKTLEN))
1999-12-12 14:11:15 +03:00
return NULL;
2000-05-03 15:12:03 +04:00
ic = (struct icmp *)fin->fin_dp;
1999-12-12 14:11:15 +03:00
type = ic->icmp_type;
/*
* If it's not an error type, then return
*/
if ((type != ICMP_UNREACH) && (type != ICMP_SOURCEQUENCH) &&
(type != ICMP_REDIRECT) && (type != ICMP_TIMXCEED) &&
(type != ICMP_PARAMPROB))
return NULL;
2000-05-03 15:12:03 +04:00
oip = (ip_t *)((char *)ic + ICMPERR_ICMPHLEN);
if (fin->fin_plen < ICMPERR_MAXPKTLEN + ((oip->ip_hl - 5) << 2))
1999-12-12 14:11:15 +03:00
return NULL;
if (oip->ip_p == IPPROTO_ICMP) {
icmp = (icmphdr_t *)((char *)oip + (oip->ip_hl << 2));
/*
* a ICMP error can only be generated as a result of an
* ICMP query, not as the response on an ICMP error
*
* XXX theoretically ICMP_ECHOREP and the other reply's are
* ICMP query's as well, but adding them here seems strange XXX
*/
1999-12-28 10:14:53 +03:00
if ((icmp->icmp_type != ICMP_ECHO) &&
(icmp->icmp_type != ICMP_TSTAMP) &&
(icmp->icmp_type != ICMP_IREQ) &&
2000-02-02 00:29:15 +03:00
(icmp->icmp_type != ICMP_MASKREQ))
1999-12-12 14:11:15 +03:00
return NULL;
2000-02-02 00:29:15 +03:00
/*
1999-12-12 14:11:15 +03:00
* perform a lookup of the ICMP packet in the state table
*/
hv = (pr = oip->ip_p);
2000-05-03 15:12:03 +04:00
src.in4 = oip->ip_src;
hv += src.in4.s_addr;
dst.in4 = oip->ip_dst;
hv += dst.in4.s_addr;
hv += icmp->icmp_id;
hv += icmp->icmp_seq;
1999-12-12 14:11:15 +03:00
hv %= fr_statesize;
oip->ip_len = ntohs(oip->ip_len);
fr_makefrip(oip->ip_hl << 2, oip, &ofin);
oip->ip_len = htons(oip->ip_len);
ofin.fin_ifp = fin->fin_ifp;
ofin.fin_out = !fin->fin_out;
ofin.fin_mp = NULL; /* if dereferenced, panic XXX */
1999-12-28 10:14:53 +03:00
1999-12-12 14:11:15 +03:00
READ_ENTER(&ipf_state);
2000-05-03 15:12:03 +04:00
for (isp = &ips_table[hv]; (is = *isp); isp = &is->is_hnext)
if ((is->is_p == pr) && (is->is_v == 4) &&
fr_matchsrcdst(is, src, dst, &ofin, NULL) &&
fr_matchicmpqueryreply(is->is_v, is, icmp)) {
ips_stats.iss_hits++;
is->is_pkts++;
is->is_bytes += ip->ip_len;
fr = is->is_rule;
RWLOCK_EXIT(&ipf_state);
return fr;
1999-12-12 14:11:15 +03:00
}
RWLOCK_EXIT(&ipf_state);
return NULL;
};
if ((oip->ip_p != IPPROTO_TCP) && (oip->ip_p != IPPROTO_UDP))
return NULL;
tcp = (tcphdr_t *)((char *)oip + (oip->ip_hl << 2));
dport = tcp->th_dport;
sport = tcp->th_sport;
hv = (pr = oip->ip_p);
2000-05-03 15:12:03 +04:00
src.in4 = oip->ip_src;
hv += src.in4.s_addr;
dst.in4 = oip->ip_dst;
hv += dst.in4.s_addr;
1999-12-12 14:11:15 +03:00
hv += dport;
hv += sport;
hv %= fr_statesize;
/*
* we make an fin entry to be able to feed it to
* matchsrcdst note that not all fields are encessary
* but this is the cleanest way. Note further we fill
* in fin_mp such that if someone uses it we'll get
* a kernel panic. fr_matchsrcdst does not use this.
*
* watch out here, as ip is in host order and oip in network
* order. Any change we make must be undone afterwards.
*/
savelen = oip->ip_len;
oip->ip_len = ip->ip_len - (ip->ip_hl << 2) - ICMPERR_ICMPHLEN;
fr_makefrip(oip->ip_hl << 2, oip, &ofin);
oip->ip_len = savelen;
ofin.fin_ifp = fin->fin_ifp;
ofin.fin_out = !fin->fin_out;
ofin.fin_mp = NULL; /* if dereferenced, panic XXX */
READ_ENTER(&ipf_state);
2000-05-03 15:12:03 +04:00
for (isp = &ips_table[hv]; (is = *isp); isp = &is->is_hnext) {
1999-12-12 14:11:15 +03:00
/*
* Only allow this icmp though if the
* encapsulated packet was allowed through the
* other way around. Note that the minimal amount
* of info present does not allow for checking against
* tcp internals such as seq and ack numbers.
*/
2000-05-03 15:12:03 +04:00
if ((is->is_p == pr) && (is->is_v == 4) &&
1999-12-12 14:11:15 +03:00
fr_matchsrcdst(is, src, dst, &ofin, tcp)) {
fr = is->is_rule;
ips_stats.iss_hits++;
/*
* we must swap src and dst here because the icmp
* comes the other way around
*/
is->is_pkts++;
2000-05-03 15:12:03 +04:00
is->is_bytes += fin->fin_plen;
1999-12-12 14:11:15 +03:00
/*
* we deliberately do not touch the timeouts
* for the accompanying state table entry.
* It remains to be seen if that is correct. XXX
*/
RWLOCK_EXIT(&ipf_state);
return fr;
}
}
RWLOCK_EXIT(&ipf_state);
return NULL;
}
/*
* Check if a packet has a registered state.
*/
1999-12-12 14:11:15 +03:00
frentry_t *fr_checkstate(ip, fin)
ip_t *ip;
fr_info_t *fin;
{
2000-05-03 15:12:03 +04:00
union i6addr dst, src;
register ipstate_t *is, **isp;
register u_char pr;
u_int hv, hvm, hlen, tryagain, pass, v;
struct icmp *ic;
1999-12-12 14:11:15 +03:00
frentry_t *fr;
tcphdr_t *tcp;
2000-05-03 15:12:03 +04:00
if (fr_state_lock || (fin->fin_off & IP_OFFMASK) ||
(fin->fin_fi.fi_fl & FI_SHORT))
1999-12-12 14:11:15 +03:00
return NULL;
1999-12-12 14:11:15 +03:00
is = NULL;
hlen = fin->fin_hlen;
tcp = (tcphdr_t *)((char *)ip + hlen);
ic = (struct icmp *)tcp;
2000-05-03 15:12:03 +04:00
hv = (pr = fin->fin_fi.fi_p);
src = fin->fin_fi.fi_src;
dst = fin->fin_fi.fi_dst;
hv += src.in4.s_addr;
hv += dst.in4.s_addr;
/*
* Search the hash table for matching packet header info.
*/
2000-05-03 15:12:03 +04:00
v = fin->fin_fi.fi_v;
switch (fin->fin_fi.fi_p)
{
2000-05-03 15:12:03 +04:00
#ifdef USE_INET6
case IPPROTO_ICMPV6 :
if (v == 6) {
if (fin->fin_out)
hv -= dst.in4.s_addr;
else
hv -= src.in4.s_addr;
if ((ic->icmp_type == ICMP6_ECHO_REQUEST) ||
(ic->icmp_type == ICMP6_ECHO_REPLY)) {
hv += ic->icmp_id;
hv += ic->icmp_seq;
}
}
#endif
case IPPROTO_ICMP :
2000-05-03 15:12:03 +04:00
if (v == 4) {
1999-12-28 10:14:53 +03:00
hv += ic->icmp_id;
hv += ic->icmp_seq;
}
1999-12-12 14:11:15 +03:00
hv %= fr_statesize;
1998-07-12 19:23:59 +04:00
READ_ENTER(&ipf_state);
2000-05-03 15:12:03 +04:00
for (isp = &ips_table[hv]; (is = *isp); isp = &is->is_hnext) {
if ((is->is_p == pr) && (is->is_v == v) &&
fr_matchsrcdst(is, src, dst, fin, NULL) &&
fr_matchicmpqueryreply(v, is, ic)) {
is->is_age = fr_icmptimeout;
1999-12-12 14:11:15 +03:00
break;
}
2000-05-03 15:12:03 +04:00
}
1999-12-12 14:11:15 +03:00
if (is != NULL)
break;
1998-07-12 19:23:59 +04:00
RWLOCK_EXIT(&ipf_state);
1999-12-12 14:11:15 +03:00
/*
* No matching icmp state entry. Perhaps this is a
* response to another state entry.
*/
2000-05-03 15:12:03 +04:00
#ifdef USE_INET6
if (v == 6)
fr = fr_checkicmp6matchingstate((ip6_t *)ip, fin);
else
#endif
fr = fr_checkicmpmatchingstate(ip, fin);
1999-12-12 14:11:15 +03:00
if (fr)
return fr;
break;
case IPPROTO_TCP :
{
2000-03-30 16:51:13 +04:00
u_short dport = tcp->th_dport, sport = tcp->th_sport;
1999-12-12 14:11:15 +03:00
tryagain = 0;
retry_tcp:
hvm = hv % fr_statesize;
1998-07-12 19:23:59 +04:00
WRITE_ENTER(&ipf_state);
1999-12-12 14:11:15 +03:00
for (isp = &ips_table[hvm]; (is = *isp);
2000-05-03 15:12:03 +04:00
isp = &is->is_hnext)
if ((is->is_p == pr) && (is->is_v == v) &&
1999-12-12 14:11:15 +03:00
fr_matchsrcdst(is, src, dst, fin, tcp)) {
2000-05-03 15:12:03 +04:00
if (fr_tcpstate(is, fin, ip, tcp))
1999-12-12 14:11:15 +03:00
break;
is = NULL;
break;
}
1999-12-12 14:11:15 +03:00
if (is != NULL)
break;
1998-07-12 19:23:59 +04:00
RWLOCK_EXIT(&ipf_state);
1999-12-12 14:11:15 +03:00
hv += dport;
hv += sport;
if (tryagain == 0) {
tryagain = 1;
goto retry_tcp;
}
break;
}
case IPPROTO_UDP :
{
2000-03-30 16:51:13 +04:00
u_short dport = tcp->th_dport, sport = tcp->th_sport;
1999-12-12 14:11:15 +03:00
tryagain = 0;
retry_udp:
hvm = hv % fr_statesize;
/*
* Nothing else to match on but ports. and IP#'s
*/
1998-07-12 19:23:59 +04:00
READ_ENTER(&ipf_state);
2000-05-03 15:12:03 +04:00
for (is = ips_table[hvm]; is; is = is->is_hnext)
if ((is->is_p == pr) && (is->is_v == v) &&
1999-12-12 14:11:15 +03:00
fr_matchsrcdst(is, src, dst, fin, tcp)) {
is->is_age = fr_udptimeout;
1999-12-12 14:11:15 +03:00
break;
}
1999-12-12 14:11:15 +03:00
if (is != NULL)
break;
1998-07-12 19:23:59 +04:00
RWLOCK_EXIT(&ipf_state);
1999-12-12 14:11:15 +03:00
hv += dport;
hv += sport;
if (tryagain == 0) {
tryagain = 1;
goto retry_udp;
}
break;
}
default :
break;
}
1999-12-12 14:11:15 +03:00
if (is == NULL) {
2000-05-03 15:12:03 +04:00
ATOMIC_INCL(ips_stats.iss_miss);
1999-12-12 14:11:15 +03:00
return NULL;
}
2000-05-03 15:12:03 +04:00
MUTEX_ENTER(&is->is_lock);
is->is_bytes += fin->fin_plen;
1999-12-12 14:11:15 +03:00
ips_stats.iss_hits++;
is->is_pkts++;
2000-05-03 15:12:03 +04:00
MUTEX_EXIT(&is->is_lock);
1999-12-12 14:11:15 +03:00
fr = is->is_rule;
fin->fin_fr = fr;
pass = is->is_pass;
#ifndef _KERNEL
if (tcp->th_flags & TCP_CLOSE)
fr_delstate(is);
#endif
1999-12-12 14:11:15 +03:00
RWLOCK_EXIT(&ipf_state);
if (fin->fin_fi.fi_fl & FI_FRAG)
ipfr_newfrag(ip, fin, pass ^ FR_KEEPSTATE);
return fr;
}
2000-05-03 15:12:03 +04:00
void ip_statesync(ifp)
void *ifp;
{
register ipstate_t *is;
WRITE_ENTER(&ipf_state);
for (is = ips_list; is; is = is->is_next) {
if (is->is_ifpin == ifp) {
is->is_ifpin = GETUNIT(is->is_ifname[0], is->is_v);
if (!is->is_ifpin)
is->is_ifpin = (void *)-1;
}
if (is->is_ifpout == ifp) {
is->is_ifpout = GETUNIT(is->is_ifname[1], is->is_v);
if (!is->is_ifpout)
is->is_ifpout = (void *)-1;
}
}
RWLOCK_EXIT(&ipf_state);
}
1999-12-12 14:11:15 +03:00
static void fr_delstate(is)
ipstate_t *is;
{
frentry_t *fr;
2000-05-03 15:12:03 +04:00
if (is->is_next)
is->is_next->is_pnext = is->is_pnext;
*is->is_pnext = is->is_next;
if (is->is_hnext)
is->is_hnext->is_phnext = is->is_phnext;
*is->is_phnext = is->is_hnext;
if (ips_table[is->is_hv] == NULL)
ips_stats.iss_inuse--;
1999-12-12 14:11:15 +03:00
fr = is->is_rule;
if (fr != NULL) {
2000-05-03 15:12:03 +04:00
ATOMIC_DEC32(fr->fr_ref);
1999-12-12 14:11:15 +03:00
if (fr->fr_ref == 0)
KFREE(fr);
}
2000-05-03 15:12:03 +04:00
#ifdef _KERNEL
MUTEX_DESTROY(&is->is_lock);
#endif
1999-12-12 14:11:15 +03:00
KFREE(is);
2000-05-03 15:12:03 +04:00
ips_num--;
}
/*
* Free memory in use by all state info. kept.
*/
void fr_stateunload()
{
2000-05-03 15:12:03 +04:00
register ipstate_t *is;
1998-07-12 19:23:59 +04:00
WRITE_ENTER(&ipf_state);
2000-05-03 15:12:03 +04:00
while ((is = ips_list))
fr_delstate(is);
1999-12-12 14:11:15 +03:00
ips_stats.iss_inuse = 0;
1998-11-22 18:17:18 +03:00
ips_num = 0;
1998-07-12 19:23:59 +04:00
RWLOCK_EXIT(&ipf_state);
1999-12-12 14:11:15 +03:00
KFREES(ips_table, fr_statesize * sizeof(ipstate_t *));
ips_table = NULL;
}
/*
* Slowly expire held state for thingslike UDP and ICMP. Timeouts are set
* in expectation of this being called twice per second.
*/
void fr_timeoutstate()
{
2000-05-03 15:12:03 +04:00
register ipstate_t *is, **isp;
1997-07-05 09:38:14 +04:00
#if defined(_KERNEL) && !SOLARIS
int s;
1997-07-05 09:38:14 +04:00
#endif
SPL_NET(s);
1998-07-12 19:23:59 +04:00
WRITE_ENTER(&ipf_state);
2000-05-03 15:12:03 +04:00
for (isp = &ips_list; (is = *isp); )
if (is->is_age && !--is->is_age) {
if (is->is_p == IPPROTO_TCP)
ips_stats.iss_fin++;
else
ips_stats.iss_expire++;
1997-05-25 16:40:11 +04:00
#ifdef IPFILTER_LOG
2000-05-03 15:12:03 +04:00
ipstate_log(is, ISL_EXPIRE);
1997-05-25 16:40:11 +04:00
#endif
2000-05-03 15:12:03 +04:00
fr_delstate(is);
} else
isp = &is->is_next;
RWLOCK_EXIT(&ipf_state);
2000-02-02 00:29:15 +03:00
SPL_X(s);
1999-12-12 14:11:15 +03:00
if (fr_state_doflush) {
(void) fr_state_flush(1);
fr_state_doflush = 0;
}
}
/*
* Original idea freom Pradeep Krishnan for use primarily with NAT code.
* (pkrishna@netcom.com)
*/
2000-05-03 15:12:03 +04:00
void fr_tcp_age(age, state, fin, dir)
u_long *age;
u_char *state;
fr_info_t *fin;
int dir;
{
tcphdr_t *tcp = (tcphdr_t *)fin->fin_dp;
u_char flags = tcp->th_flags;
int dlen, ostate;
ostate = state[1 - dir];
2000-05-03 15:12:03 +04:00
dlen = fin->fin_plen - fin->fin_hlen - (tcp->th_off << 2);
if (flags & TH_RST) {
if (!(tcp->th_flags & TH_PUSH) && !dlen) {
*age = fr_tcpclosed;
state[dir] = TCPS_CLOSED;
} else {
*age = fr_tcpclosewait;
state[dir] = TCPS_CLOSE_WAIT;
}
return;
}
*age = fr_tcptimeout; /* 1 min */
switch(state[dir])
{
case TCPS_CLOSED:
1999-12-12 14:11:15 +03:00
if ((flags & (TH_FIN|TH_SYN|TH_RST|TH_ACK)) == TH_ACK) {
state[dir] = TCPS_ESTABLISHED;
*age = fr_tcpidletimeout;
}
case TCPS_FIN_WAIT_2:
if ((flags & TH_OPENING) == TH_OPENING)
state[dir] = TCPS_SYN_RECEIVED;
else if (flags & TH_SYN)
state[dir] = TCPS_SYN_SENT;
break;
case TCPS_SYN_RECEIVED:
case TCPS_SYN_SENT:
if ((flags & (TH_FIN|TH_ACK)) == TH_ACK) {
state[dir] = TCPS_ESTABLISHED;
*age = fr_tcpidletimeout;
1999-12-12 14:11:15 +03:00
} else if ((flags & (TH_FIN|TH_ACK)) == (TH_FIN|TH_ACK)) {
state[dir] = TCPS_CLOSE_WAIT;
if (!(flags & TH_PUSH) && !dlen &&
ostate > TCPS_ESTABLISHED)
*age = fr_tcplastack;
else
*age = fr_tcpclosewait;
}
break;
case TCPS_ESTABLISHED:
if (flags & TH_FIN) {
state[dir] = TCPS_CLOSE_WAIT;
if (!(flags & TH_PUSH) && !dlen &&
ostate > TCPS_ESTABLISHED)
*age = fr_tcplastack;
else
*age = fr_tcpclosewait;
1999-12-12 14:11:15 +03:00
} else {
if (ostate < TCPS_CLOSE_WAIT)
*age = fr_tcpidletimeout;
}
break;
case TCPS_CLOSE_WAIT:
if ((flags & TH_FIN) && !(flags & TH_PUSH) && !dlen &&
ostate > TCPS_ESTABLISHED) {
*age = fr_tcplastack;
state[dir] = TCPS_LAST_ACK;
} else
*age = fr_tcpclosewait;
break;
case TCPS_LAST_ACK:
if (flags & TH_ACK) {
state[dir] = TCPS_FIN_WAIT_2;
if (!(flags & TH_PUSH) && !dlen &&
ostate > TCPS_ESTABLISHED)
*age = fr_tcplastack;
else {
*age = fr_tcpclosewait;
state[dir] = TCPS_CLOSE_WAIT;
}
}
break;
}
}
1997-05-25 16:40:11 +04:00
#ifdef IPFILTER_LOG
void ipstate_log(is, type)
struct ipstate *is;
1999-12-12 14:11:15 +03:00
u_int type;
1997-05-25 16:40:11 +04:00
{
struct ipslog ipsl;
void *items[1];
size_t sizes[1];
int types[1];
1997-05-25 16:40:11 +04:00
1998-11-22 18:17:18 +03:00
ipsl.isl_type = type;
1997-05-25 16:40:11 +04:00
ipsl.isl_pkts = is->is_pkts;
ipsl.isl_bytes = is->is_bytes;
ipsl.isl_src = is->is_src;
ipsl.isl_dst = is->is_dst;
ipsl.isl_p = is->is_p;
2000-05-03 15:12:03 +04:00
ipsl.isl_v = is->is_v;
1997-05-25 16:40:11 +04:00
ipsl.isl_flags = is->is_flags;
if (ipsl.isl_p == IPPROTO_TCP || ipsl.isl_p == IPPROTO_UDP) {
ipsl.isl_sport = is->is_sport;
ipsl.isl_dport = is->is_dport;
1998-11-22 18:17:18 +03:00
if (ipsl.isl_p == IPPROTO_TCP) {
ipsl.isl_state[0] = is->is_state[0];
ipsl.isl_state[1] = is->is_state[1];
}
1997-05-25 16:40:11 +04:00
} else if (ipsl.isl_p == IPPROTO_ICMP)
ipsl.isl_itype = is->is_icmp.ics_type;
else {
ipsl.isl_ps.isl_filler[0] = 0;
ipsl.isl_ps.isl_filler[1] = 0;
}
items[0] = &ipsl;
sizes[0] = sizeof(ipsl);
types[0] = 0;
1997-05-25 16:40:11 +04:00
1999-12-12 14:11:15 +03:00
(void) ipllog(IPL_LOGSTATE, NULL, items, sizes, types, 1);
1997-05-25 16:40:11 +04:00
}
#endif
2000-02-02 00:29:15 +03:00
2000-05-03 15:12:03 +04:00
#ifdef USE_INET6
frentry_t *fr_checkicmp6matchingstate(ip, fin)
ip6_t *ip;
fr_info_t *fin;
2000-02-02 00:29:15 +03:00
{
2000-05-03 15:12:03 +04:00
register ipstate_t *is, **isp;
register u_short sport, dport;
register u_char pr;
struct icmp6_hdr *ic, *oic;
union i6addr dst, src;
u_short savelen;
fr_info_t ofin;
tcphdr_t *tcp;
frentry_t *fr;
ip6_t *oip;
int type;
u_int hv;
2000-02-02 00:29:15 +03:00
2000-05-03 15:12:03 +04:00
/*
* Does it at least have the return (basic) IP header ?
* Only a basic IP header (no options) should be with
* an ICMP error header.
*/
if ((fin->fin_v != 6) || (fin->fin_plen < ICMP6ERR_MINPKTLEN))
return NULL;
ic = (struct icmp6_hdr *)fin->fin_dp;
type = ic->icmp6_type;
/*
* If it's not an error type, then return
*/
if ((type != ICMP6_DST_UNREACH) && (type != ICMP6_PACKET_TOO_BIG) &&
(type != ICMP6_TIME_EXCEEDED) && (type != ICMP6_PARAM_PROB))
return NULL;
oip = (ip6_t *)((char *)ic + ICMPERR_ICMPHLEN);
if (fin->fin_plen < sizeof(*oip))
return NULL;
if (oip->ip6_nxt == IPPROTO_ICMPV6) {
oic = (struct icmp6_hdr *)(oip + 1);
/*
* a ICMP error can only be generated as a result of an
* ICMP query, not as the response on an ICMP error
*
* XXX theoretically ICMP_ECHOREP and the other reply's are
* ICMP query's as well, but adding them here seems strange XXX
*/
if (!(oic->icmp6_type & ICMP6_INFOMSG_MASK))
return NULL;
/*
* perform a lookup of the ICMP packet in the state table
*/
hv = (pr = oip->ip6_nxt);
src.in6 = oip->ip6_src;
hv += src.in4.s_addr;
dst.in6 = oip->ip6_dst;
hv += dst.in4.s_addr;
hv += oic->icmp6_id;
hv += oic->icmp6_seq;
hv %= fr_statesize;
oip->ip6_plen = ntohs(oip->ip6_plen);
fr_makefrip(sizeof(*oip), (ip_t *)oip, &ofin);
oip->ip6_plen = htons(oip->ip6_plen);
ofin.fin_ifp = fin->fin_ifp;
ofin.fin_out = !fin->fin_out;
ofin.fin_mp = NULL; /* if dereferenced, panic XXX */
READ_ENTER(&ipf_state);
for (isp = &ips_table[hv]; (is = *isp); isp = &is->is_hnext)
if ((is->is_p == pr) &&
(oic->icmp6_id == is->is_icmp.ics_id) &&
(oic->icmp6_seq == is->is_icmp.ics_seq) &&
fr_matchsrcdst(is, src, dst, &ofin, NULL)) {
/*
* in the state table ICMP query's are stored
* with the type of the corresponding ICMP
* response. Correct here
*/
if (((is->is_type == ICMP6_ECHO_REPLY) &&
(oic->icmp6_type == ICMP6_ECHO_REQUEST)) ||
(is->is_type - 1 == oic->icmp6_type )) {
ips_stats.iss_hits++;
is->is_pkts++;
is->is_bytes += fin->fin_plen;
return is->is_rule;
}
}
RWLOCK_EXIT(&ipf_state);
return NULL;
};
if ((oip->ip6_nxt != IPPROTO_TCP) && (oip->ip6_nxt != IPPROTO_UDP))
return NULL;
tcp = (tcphdr_t *)(oip + 1);
dport = tcp->th_dport;
sport = tcp->th_sport;
hv = (pr = oip->ip6_nxt);
src.in6 = oip->ip6_src;
hv += src.in4.s_addr;
dst.in6 = oip->ip6_dst;
hv += dst.in4.s_addr;
hv += dport;
hv += sport;
hv %= fr_statesize;
/*
* we make an fin entry to be able to feed it to
* matchsrcdst note that not all fields are encessary
* but this is the cleanest way. Note further we fill
* in fin_mp such that if someone uses it we'll get
* a kernel panic. fr_matchsrcdst does not use this.
*
* watch out here, as ip is in host order and oip in network
* order. Any change we make must be undone afterwards.
*/
savelen = oip->ip6_plen;
oip->ip6_plen = ip->ip6_plen - sizeof(*ip) - ICMPERR_ICMPHLEN;
ofin.fin_v = 6;
fr_makefrip(sizeof(*oip), (ip_t *)oip, &ofin);
oip->ip6_plen = savelen;
ofin.fin_ifp = fin->fin_ifp;
ofin.fin_out = !fin->fin_out;
ofin.fin_mp = NULL; /* if dereferenced, panic XXX */
READ_ENTER(&ipf_state);
for (isp = &ips_table[hv]; (is = *isp); isp = &is->is_hnext) {
/*
* Only allow this icmp though if the
* encapsulated packet was allowed through the
* other way around. Note that the minimal amount
* of info present does not allow for checking against
* tcp internals such as seq and ack numbers.
*/
if ((is->is_p == pr) && (is->is_v == 6) &&
fr_matchsrcdst(is, src, dst, &ofin, tcp)) {
fr = is->is_rule;
ips_stats.iss_hits++;
/*
* we must swap src and dst here because the icmp
* comes the other way around
*/
is->is_pkts++;
is->is_bytes += fin->fin_plen;
/*
* we deliberately do not touch the timeouts
* for the accompanying state table entry.
* It remains to be seen if that is correct. XXX
*/
RWLOCK_EXIT(&ipf_state);
return fr;
2000-02-02 00:29:15 +03:00
}
2000-05-03 15:12:03 +04:00
}
2000-02-02 00:29:15 +03:00
RWLOCK_EXIT(&ipf_state);
2000-05-03 15:12:03 +04:00
return NULL;
2000-02-02 00:29:15 +03:00
}
2000-05-03 15:12:03 +04:00
#endif