aboutsummaryrefslogtreecommitdiff
path: root/contrib/ipfilter/ip_frag.c
diff options
context:
space:
mode:
Diffstat (limited to 'contrib/ipfilter/ip_frag.c')
-rw-r--r--contrib/ipfilter/ip_frag.c555
1 files changed, 0 insertions, 555 deletions
diff --git a/contrib/ipfilter/ip_frag.c b/contrib/ipfilter/ip_frag.c
deleted file mode 100644
index 556478d51d0b..000000000000
--- a/contrib/ipfilter/ip_frag.c
+++ /dev/null
@@ -1,555 +0,0 @@
-/*
- * Copyright (C) 1993-2000 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)
-static const char sccsid[] = "@(#)ip_frag.c 1.11 3/24/96 (C) 1993-2000 Darren Reed";
-static const char rcsid[] = "@(#)$Id: ip_frag.c,v 2.10.2.7 2000/11/27 10:26:56 darrenr Exp $";
-#endif
-
-#if defined(KERNEL) && !defined(_KERNEL)
-# define _KERNEL
-#endif
-
-#include <sys/errno.h>
-#include <sys/types.h>
-#include <sys/param.h>
-#include <sys/time.h>
-#include <sys/file.h>
-#if !defined(_KERNEL) && !defined(KERNEL)
-# include <stdio.h>
-# include <string.h>
-# include <stdlib.h>
-#endif
-#if (defined(KERNEL) || defined(_KERNEL)) && (__FreeBSD_version >= 220000)
-# include <sys/filio.h>
-# include <sys/fcntl.h>
-#else
-# include <sys/ioctl.h>
-#endif
-#include <sys/uio.h>
-#ifndef linux
-# include <sys/protosw.h>
-#endif
-#include <sys/socket.h>
-#if defined(_KERNEL) && !defined(linux)
-# include <sys/systm.h>
-#endif
-#if !defined(__SVR4) && !defined(__svr4__)
-# if defined(_KERNEL) && !defined(__sgi)
-# include <sys/kernel.h>
-# endif
-# ifndef linux
-# include <sys/mbuf.h>
-# endif
-#else
-# include <sys/byteorder.h>
-# ifdef _KERNEL
-# include <sys/dditypes.h>
-# endif
-# include <sys/stream.h>
-# include <sys/kmem.h>
-#endif
-#include <net/if.h>
-#ifdef sun
-# include <net/af.h>
-#endif
-#include <net/route.h>
-#include <netinet/in.h>
-#include <netinet/in_systm.h>
-#include <netinet/ip.h>
-#ifndef linux
-# include <netinet/ip_var.h>
-#endif
-#include <netinet/tcp.h>
-#include <netinet/udp.h>
-#include <netinet/ip_icmp.h>
-#include "netinet/ip_compat.h"
-#include <netinet/tcpip.h>
-#include "netinet/ip_fil.h"
-#include "netinet/ip_proxy.h"
-#include "netinet/ip_nat.h"
-#include "netinet/ip_frag.h"
-#include "netinet/ip_state.h"
-#include "netinet/ip_auth.h"
-#if (__FreeBSD_version >= 300000)
-# include <sys/malloc.h>
-# if (defined(KERNEL) || defined(_KERNEL))
-# ifndef IPFILTER_LKM
-# include <sys/libkern.h>
-# include <sys/systm.h>
-# endif
-extern struct callout_handle ipfr_slowtimer_ch;
-# endif
-#endif
-#if defined(__NetBSD__) && (__NetBSD_Version__ >= 104230000)
-# include <sys/callout.h>
-extern struct callout ipfr_slowtimer_ch;
-#endif
-
-
-static ipfr_t *ipfr_heads[IPFT_SIZE];
-static ipfr_t *ipfr_nattab[IPFT_SIZE];
-static ipfrstat_t ipfr_stats;
-static int ipfr_inuse = 0;
-
-int fr_ipfrttl = 120; /* 60 seconds */
-int fr_frag_lock = 0;
-
-#ifdef _KERNEL
-# if SOLARIS2 >= 7
-extern timeout_id_t ipfr_timer_id;
-# else
-extern int ipfr_timer_id;
-# endif
-#endif
-#if (SOLARIS || defined(__sgi)) && defined(_KERNEL)
-extern KRWLOCK_T ipf_frag, ipf_natfrag, ipf_nat, ipf_mutex;
-# if SOLARIS
-extern KRWLOCK_T ipf_solaris;
-# else
-KRWLOCK_T ipf_solaris;
-# endif
-extern kmutex_t ipf_rw;
-#endif
-
-
-static ipfr_t *ipfr_new __P((ip_t *, fr_info_t *, u_int, ipfr_t **));
-static ipfr_t *ipfr_lookup __P((ip_t *, fr_info_t *, ipfr_t **));
-static void ipfr_delete __P((ipfr_t *));
-
-
-ipfrstat_t *ipfr_fragstats()
-{
- ipfr_stats.ifs_table = ipfr_heads;
- ipfr_stats.ifs_nattab = ipfr_nattab;
- ipfr_stats.ifs_inuse = ipfr_inuse;
- return &ipfr_stats;
-}
-
-
-/*
- * add a new entry to the fragment cache, registering it as having come
- * through this box, with the result of the filter operation.
- */
-static ipfr_t *ipfr_new(ip, fin, pass, table)
-ip_t *ip;
-fr_info_t *fin;
-u_int pass;
-ipfr_t *table[];
-{
- ipfr_t **fp, *fra, frag;
- u_int idx;
-
- if (ipfr_inuse >= IPFT_SIZE)
- return NULL;
-
- frag.ipfr_p = ip->ip_p;
- idx = ip->ip_p;
- frag.ipfr_id = ip->ip_id;
- idx += ip->ip_id;
- frag.ipfr_tos = ip->ip_tos;
- frag.ipfr_src.s_addr = ip->ip_src.s_addr;
- idx += ip->ip_src.s_addr;
- frag.ipfr_dst.s_addr = ip->ip_dst.s_addr;
- idx += ip->ip_dst.s_addr;
- frag.ipfr_ifp = fin->fin_ifp;
- idx *= 127;
- idx %= IPFT_SIZE;
-
- /*
- * first, make sure it isn't already there...
- */
- for (fp = &table[idx]; (fra = *fp); fp = &fra->ipfr_next)
- if (!bcmp((char *)&frag.ipfr_src, (char *)&fra->ipfr_src,
- IPFR_CMPSZ)) {
- ATOMIC_INCL(ipfr_stats.ifs_exists);
- return NULL;
- }
-
- /*
- * allocate some memory, if possible, if not, just record that we
- * failed to do so.
- */
- KMALLOC(fra, ipfr_t *);
- if (fra == NULL) {
- ATOMIC_INCL(ipfr_stats.ifs_nomem);
- return NULL;
- }
-
- if ((fra->ipfr_rule = fin->fin_fr) != NULL) {
- ATOMIC_INC32(fin->fin_fr->fr_ref);
- }
-
-
- /*
- * Instert the fragment into the fragment table, copy the struct used
- * in the search using bcopy rather than reassign each field.
- * Set the ttl to the default and mask out logging from "pass"
- */
- if ((fra->ipfr_next = table[idx]))
- table[idx]->ipfr_prev = fra;
- fra->ipfr_prev = NULL;
- fra->ipfr_data = NULL;
- table[idx] = fra;
- bcopy((char *)&frag.ipfr_src, (char *)&fra->ipfr_src, IPFR_CMPSZ);
- fra->ipfr_ttl = fr_ipfrttl;
- /*
- * Compute the offset of the expected start of the next packet.
- */
- fra->ipfr_off = (ip->ip_off & IP_OFFMASK) + (fin->fin_dlen >> 3);
- ATOMIC_INCL(ipfr_stats.ifs_new);
- ATOMIC_INC32(ipfr_inuse);
- return fra;
-}
-
-
-int ipfr_newfrag(ip, fin, pass)
-ip_t *ip;
-fr_info_t *fin;
-u_int pass;
-{
- ipfr_t *ipf;
-
- if ((ip->ip_v != 4) || (fr_frag_lock))
- return -1;
- WRITE_ENTER(&ipf_frag);
- ipf = ipfr_new(ip, fin, pass, ipfr_heads);
- RWLOCK_EXIT(&ipf_frag);
- return ipf ? 0 : -1;
-}
-
-
-int ipfr_nat_newfrag(ip, fin, pass, nat)
-ip_t *ip;
-fr_info_t *fin;
-u_int pass;
-nat_t *nat;
-{
- ipfr_t *ipf;
-
- if ((ip->ip_v != 4) || (fr_frag_lock))
- return -1;
- WRITE_ENTER(&ipf_natfrag);
- ipf = ipfr_new(ip, fin, pass, ipfr_nattab);
- if (ipf != NULL) {
- ipf->ipfr_data = nat;
- nat->nat_data = ipf;
- }
- RWLOCK_EXIT(&ipf_natfrag);
- return ipf ? 0 : -1;
-}
-
-
-/*
- * check the fragment cache to see if there is already a record of this packet
- * with its filter result known.
- */
-static ipfr_t *ipfr_lookup(ip, fin, table)
-ip_t *ip;
-fr_info_t *fin;
-ipfr_t *table[];
-{
- ipfr_t *f, frag;
- u_int idx;
-
- /*
- * For fragments, we record protocol, packet id, TOS and both IP#'s
- * (these should all be the same for all fragments of a packet).
- *
- * build up a hash value to index the table with.
- */
- frag.ipfr_p = ip->ip_p;
- idx = ip->ip_p;
- frag.ipfr_id = ip->ip_id;
- idx += ip->ip_id;
- frag.ipfr_tos = ip->ip_tos;
- frag.ipfr_src.s_addr = ip->ip_src.s_addr;
- idx += ip->ip_src.s_addr;
- frag.ipfr_dst.s_addr = ip->ip_dst.s_addr;
- idx += ip->ip_dst.s_addr;
- frag.ipfr_ifp = fin->fin_ifp;
- idx *= 127;
- idx %= IPFT_SIZE;
-
- /*
- * check the table, careful to only compare the right amount of data
- */
- for (f = table[idx]; f; f = f->ipfr_next)
- if (!bcmp((char *)&frag.ipfr_src, (char *)&f->ipfr_src,
- IPFR_CMPSZ)) {
- u_short atoff, off;
-
- if (f != table[idx]) {
- /*
- * move fragment info. to the top of the list
- * to speed up searches.
- */
- if ((f->ipfr_prev->ipfr_next = f->ipfr_next))
- f->ipfr_next->ipfr_prev = f->ipfr_prev;
- f->ipfr_next = table[idx];
- table[idx]->ipfr_prev = f;
- f->ipfr_prev = NULL;
- table[idx] = f;
- }
- off = ip->ip_off & IP_OFFMASK;
- atoff = off + (fin->fin_dlen >> 3);
- /*
- * If we've follwed the fragments, and this is the
- * last (in order), shrink expiration time.
- */
- if (off == f->ipfr_off) {
- if (!(ip->ip_off & IP_MF))
- f->ipfr_ttl = 1;
- else
- f->ipfr_off = atoff;
- }
- ATOMIC_INCL(ipfr_stats.ifs_hits);
- return f;
- }
- return NULL;
-}
-
-
-/*
- * functional interface for NAT lookups of the NAT fragment cache
- */
-nat_t *ipfr_nat_knownfrag(ip, fin)
-ip_t *ip;
-fr_info_t *fin;
-{
- nat_t *nat;
- ipfr_t *ipf;
-
- if ((ip->ip_v != 4) || (fr_frag_lock))
- return NULL;
- READ_ENTER(&ipf_natfrag);
- ipf = ipfr_lookup(ip, fin, ipfr_nattab);
- if (ipf != NULL) {
- nat = ipf->ipfr_data;
- /*
- * This is the last fragment for this packet.
- */
- if ((ipf->ipfr_ttl == 1) && (nat != NULL)) {
- nat->nat_data = NULL;
- ipf->ipfr_data = NULL;
- }
- } else
- nat = NULL;
- RWLOCK_EXIT(&ipf_natfrag);
- return nat;
-}
-
-
-/*
- * functional interface for normal lookups of the fragment cache
- */
-frentry_t *ipfr_knownfrag(ip, fin)
-ip_t *ip;
-fr_info_t *fin;
-{
- frentry_t *fr = NULL;
- ipfr_t *fra;
-
- if ((ip->ip_v != 4) || (fr_frag_lock))
- return NULL;
- READ_ENTER(&ipf_frag);
- fra = ipfr_lookup(ip, fin, ipfr_heads);
- if (fra != NULL)
- fr = fra->ipfr_rule;
- RWLOCK_EXIT(&ipf_frag);
- return fr;
-}
-
-
-/*
- * forget any references to this external object.
- */
-void ipfr_forget(nat)
-void *nat;
-{
- ipfr_t *fr;
- int idx;
-
- WRITE_ENTER(&ipf_natfrag);
- for (idx = IPFT_SIZE - 1; idx >= 0; idx--)
- for (fr = ipfr_heads[idx]; fr; fr = fr->ipfr_next)
- if (fr->ipfr_data == nat)
- fr->ipfr_data = NULL;
-
- RWLOCK_EXIT(&ipf_natfrag);
-}
-
-
-static void ipfr_delete(fra)
-ipfr_t *fra;
-{
- frentry_t *fr;
-
- fr = fra->ipfr_rule;
- if (fr != NULL) {
- ATOMIC_DEC32(fr->fr_ref);
- if (fr->fr_ref == 0)
- KFREE(fr);
- }
- if (fra->ipfr_prev)
- fra->ipfr_prev->ipfr_next = fra->ipfr_next;
- if (fra->ipfr_next)
- fra->ipfr_next->ipfr_prev = fra->ipfr_prev;
- KFREE(fra);
-}
-
-
-/*
- * Free memory in use by fragment state info. kept.
- */
-void ipfr_unload()
-{
- ipfr_t **fp, *fra;
- nat_t *nat;
- int idx;
-
- WRITE_ENTER(&ipf_frag);
- for (idx = IPFT_SIZE - 1; idx >= 0; idx--)
- for (fp = &ipfr_heads[idx]; (fra = *fp); ) {
- *fp = fra->ipfr_next;
- ipfr_delete(fra);
- }
- RWLOCK_EXIT(&ipf_frag);
-
- WRITE_ENTER(&ipf_nat);
- WRITE_ENTER(&ipf_natfrag);
- for (idx = IPFT_SIZE - 1; idx >= 0; idx--)
- for (fp = &ipfr_nattab[idx]; (fra = *fp); ) {
- *fp = fra->ipfr_next;
- nat = fra->ipfr_data;
- if (nat != NULL) {
- if (nat->nat_data == fra)
- nat->nat_data = NULL;
- }
- ipfr_delete(fra);
- }
- RWLOCK_EXIT(&ipf_natfrag);
- RWLOCK_EXIT(&ipf_nat);
-}
-
-
-#ifdef _KERNEL
-void ipfr_fragexpire()
-{
- ipfr_t **fp, *fra;
- nat_t *nat;
- int idx;
-#if defined(_KERNEL)
-# if !SOLARIS
- int s;
-# endif
-#endif
-
- if (fr_frag_lock)
- return;
-
- SPL_NET(s);
- WRITE_ENTER(&ipf_frag);
-
- /*
- * Go through the entire table, looking for entries to expire,
- * decreasing the ttl by one for each entry. If it reaches 0,
- * remove it from the chain and free it.
- */
- for (idx = IPFT_SIZE - 1; idx >= 0; idx--)
- for (fp = &ipfr_heads[idx]; (fra = *fp); ) {
- --fra->ipfr_ttl;
- if (fra->ipfr_ttl == 0) {
- *fp = fra->ipfr_next;
- ipfr_delete(fra);
- ATOMIC_INCL(ipfr_stats.ifs_expire);
- ATOMIC_DEC32(ipfr_inuse);
- } else
- fp = &fra->ipfr_next;
- }
- RWLOCK_EXIT(&ipf_frag);
-
- /*
- * Same again for the NAT table, except that if the structure also
- * still points to a NAT structure, and the NAT structure points back
- * at the one to be free'd, NULL the reference from the NAT struct.
- * NOTE: We need to grab both mutex's early, and in this order so as
- * to prevent a deadlock if both try to expire at the same time.
- */
- WRITE_ENTER(&ipf_nat);
- WRITE_ENTER(&ipf_natfrag);
- for (idx = IPFT_SIZE - 1; idx >= 0; idx--)
- for (fp = &ipfr_nattab[idx]; (fra = *fp); ) {
- --fra->ipfr_ttl;
- if (fra->ipfr_ttl == 0) {
- ATOMIC_INCL(ipfr_stats.ifs_expire);
- ATOMIC_DEC32(ipfr_inuse);
- nat = fra->ipfr_data;
- if (nat != NULL) {
- if (nat->nat_data == fra)
- nat->nat_data = NULL;
- }
- *fp = fra->ipfr_next;
- ipfr_delete(fra);
- } else
- fp = &fra->ipfr_next;
- }
- RWLOCK_EXIT(&ipf_natfrag);
- RWLOCK_EXIT(&ipf_nat);
- SPL_X(s);
-}
-
-
-/*
- * Slowly expire held state for fragments. Timeouts are set * in expectation
- * of this being called twice per second.
- */
-# if (BSD >= 199306) || SOLARIS || defined(__sgi)
-# if defined(SOLARIS2) && (SOLARIS2 < 7)
-void ipfr_slowtimer()
-# else
-void ipfr_slowtimer __P((void *ptr))
-# endif
-# else
-int ipfr_slowtimer()
-# endif
-{
-#if defined(_KERNEL) && SOLARIS
- extern int fr_running;
-
- if (fr_running <= 0)
- return;
-#endif
-
- READ_ENTER(&ipf_solaris);
-#ifdef __sgi
- ipfilter_sgi_intfsync();
-#endif
-
- ipfr_fragexpire();
- fr_timeoutstate();
- ip_natexpire();
- fr_authexpire();
-# if SOLARIS
- ipfr_timer_id = timeout(ipfr_slowtimer, NULL, drv_usectohz(500000));
- RWLOCK_EXIT(&ipf_solaris);
-# else
-# if defined(__NetBSD__) && (__NetBSD_Version__ >= 104240000)
- callout_reset(&ipfr_slowtimer_ch, hz / 2, ipfr_slowtimer, NULL);
-# else
-# if (__FreeBSD_version >= 300000)
- ipfr_slowtimer_ch = timeout(ipfr_slowtimer, NULL, hz/2);
-# else
- timeout(ipfr_slowtimer, NULL, hz/2);
-# endif
-# if (BSD < 199306) && !defined(__sgi)
- return 0;
-# endif /* FreeBSD */
-# endif /* NetBSD */
-# endif /* SOLARIS */
-}
-#endif /* defined(_KERNEL) */