1 files changed, 233 insertions, 141 deletions

diff --git a/sys/contrib/ipfilter/netinet/ip_frag.c b/sys/contrib/ipfilter/netinet/ip_frag.c
index 923f685..3e0a7f3 100644
--- a/sys/contrib/ipfilter/netinet/ip_frag.c
+++ b/sys/contrib/ipfilter/netinet/ip_frag.c
@@ -1,59 +1,68 @@
 /*
- * Copyright (C) 1993-1997 by Darren Reed.
+ * 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) 1993-1995 Darren Reed";
-static const char rcsid[] = "@(#)$Id: ip_frag.c,v 2.0.2.19.2.1 1997/11/12 10:50:21 darrenr Exp $";
+static const char sccsid[] = "@(#)ip_frag.c	1.11 3/24/96 (C) 1993-2000 Darren Reed";
+static const char rcsid[] = "@(#)$Id: ip_frag.c,v 2.10.2.3 2000/05/05 15:10:23 darrenr Exp $";
 #endif
 
-#if !defined(_KERNEL) && !defined(KERNEL)
-# include <string.h>
-# include <stdlib.h>
+#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) && (__FreeBSD_version >= 220000)
-#include <sys/filio.h>
-#include <sys/fcntl.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>
+# include <sys/ioctl.h>
 #endif
 #include <sys/uio.h>
 #ifndef linux
-#include <sys/protosw.h>
+# 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>
-# include <sys/dditypes.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>
+# 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>
+# include <netinet/ip_var.h>
 #endif
 #include <netinet/tcp.h>
 #include <netinet/udp.h>
@@ -66,24 +75,51 @@ static const char rcsid[] = "@(#)$Id: ip_frag.c,v 2.0.2.19.2.1 1997/11/12 10:50:
 #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;
 
-ipfr_t	*ipfr_heads[IPFT_SIZE];
-ipfr_t	*ipfr_nattab[IPFT_SIZE];
-ipfrstat_t ipfr_stats;
-int	ipfr_inuse = 0,
-	fr_ipfrttl = 120;	/* 60 seconds */
 #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	kmutex_t	ipf_frag;
-extern	kmutex_t	ipf_natfrag;
-extern	kmutex_t	ipf_nat;
+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 *, int, ipfr_t **));
+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()
@@ -102,10 +138,10 @@ ipfrstat_t *ipfr_fragstats()
 static ipfr_t *ipfr_new(ip, fin, pass, table)
 ip_t *ip;
 fr_info_t *fin;
-int pass;
+u_int pass;
 ipfr_t *table[];
 {
-	ipfr_t	**fp, *fr, frag;
+	ipfr_t	**fp, *fra, frag;
 	u_int	idx;
 
 	frag.ipfr_p = ip->ip_p;
@@ -123,10 +159,10 @@ ipfr_t *table[];
 	/*
 	 * first, make sure it isn't already there...
 	 */
-	for (fp = &table[idx]; (fr = *fp); fp = &fr->ipfr_next)
-		if (!bcmp((char *)&frag.ipfr_src, (char *)&fr->ipfr_src,
+	for (fp = &table[idx]; (fra = *fp); fp = &fra->ipfr_next)
+		if (!bcmp((char *)&frag.ipfr_src, (char *)&fra->ipfr_src,
 			  IPFR_CMPSZ)) {
-			ipfr_stats.ifs_exists++;
+			ATOMIC_INCL(ipfr_stats.ifs_exists);
 			return NULL;
 		}
 
@@ -134,45 +170,51 @@ ipfr_t *table[];
 	 * allocate some memory, if possible, if not, just record that we
 	 * failed to do so.
 	 */
-	KMALLOC(fr, ipfr_t *, sizeof(*fr));
-	if (fr == NULL) {
-		ipfr_stats.ifs_nomem++;
+	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 ((fr->ipfr_next = table[idx]))
-		table[idx]->ipfr_prev = fr;
-	fr->ipfr_prev = NULL;
-	fr->ipfr_data = NULL;
-	table[idx] = fr;
-	bcopy((char *)&frag.ipfr_src, (char *)&fr->ipfr_src, IPFR_CMPSZ);
-	fr->ipfr_ttl = fr_ipfrttl;
-	fr->ipfr_pass = pass & ~(FR_LOGFIRST|FR_LOG);
+	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.
 	 */
-	fr->ipfr_off = (ip->ip_off & 0x1fff) + (fin->fin_dlen >> 3);
-	ipfr_stats.ifs_new++;
-	ipfr_inuse++;
-	return fr;
+	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;
-int pass;
+u_int pass;
 {
 	ipfr_t	*ipf;
 
-	MUTEX_ENTER(&ipf_frag);
+	if ((ip->ip_v != 4) || (fr_frag_lock))
+		return NULL;
+	WRITE_ENTER(&ipf_frag);
 	ipf = ipfr_new(ip, fin, pass, ipfr_heads);
-	MUTEX_EXIT(&ipf_frag);
+	RWLOCK_EXIT(&ipf_frag);
 	return ipf ? 0 : -1;
 }
 
@@ -180,17 +222,20 @@ int pass;
 int ipfr_nat_newfrag(ip, fin, pass, nat)
 ip_t *ip;
 fr_info_t *fin;
-int pass;
+u_int pass;
 nat_t *nat;
 {
 	ipfr_t	*ipf;
 
-	MUTEX_ENTER(&ipf_natfrag);
-	if ((ipf = ipfr_new(ip, fin, pass, ipfr_nattab))) {
+	if ((ip->ip_v != 4) || (fr_frag_lock))
+		return NULL;
+	WRITE_ENTER(&ipf_natfrag);
+	ipf = ipfr_new(ip, fin, pass, ipfr_nattab);
+	if (ipf != NULL) {
 		ipf->ipfr_data = nat;
 		nat->nat_data = ipf;
 	}
-	MUTEX_EXIT(&ipf_natfrag);
+	RWLOCK_EXIT(&ipf_natfrag);
 	return ipf ? 0 : -1;
 }
 
@@ -245,19 +290,19 @@ ipfr_t *table[];
 				f->ipfr_prev = NULL;
 				table[idx] = f;
 			}
-			off = ip->ip_off;
+			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 & 0x1fff) == f->ipfr_off) {
-				if (!(off & IP_MF))
+			if (off == f->ipfr_off) {
+				if (!(ip->ip_off & IP_MF))
 					f->ipfr_ttl = 1;
 				else
 					f->ipfr_off = atoff;
 			}
-			ipfr_stats.ifs_hits++;
+			ATOMIC_INCL(ipfr_stats.ifs_hits);
 			return f;
 		}
 	return NULL;
@@ -274,20 +319,22 @@ fr_info_t *fin;
 	nat_t	*nat;
 	ipfr_t	*ipf;
 
-	MUTEX_ENTER(&ipf_natfrag);
+	if ((ip->ip_v != 4) || (fr_frag_lock))
+		return NULL;
+	READ_ENTER(&ipf_natfrag);
 	ipf = ipfr_lookup(ip, fin, ipfr_nattab);
-	if (ipf) {
+	if (ipf != NULL) {
 		nat = ipf->ipfr_data;
 		/*
 		 * This is the last fragment for this packet.
 		 */
-		if (ipf->ipfr_ttl == 1) {
+		if ((ipf->ipfr_ttl == 1) && (nat != NULL)) {
 			nat->nat_data = NULL;
 			ipf->ipfr_data = NULL;
 		}
 	} else
 		nat = NULL;
-	MUTEX_EXIT(&ipf_natfrag);
+	RWLOCK_EXIT(&ipf_natfrag);
 	return nat;
 }
 
@@ -295,18 +342,21 @@ fr_info_t *fin;
 /*
  * functional interface for normal lookups of the fragment cache
  */
-int ipfr_knownfrag(ip, fin)
+frentry_t *ipfr_knownfrag(ip, fin)
 ip_t *ip;
 fr_info_t *fin;
 {
-	int	ret;
-	ipfr_t	*ipf;
+	frentry_t *fr = NULL;
+	ipfr_t	*fra;
 
-	MUTEX_ENTER(&ipf_frag);
-	ipf = ipfr_lookup(ip, fin, ipfr_heads);
-	ret = ipf ? ipf->ipfr_pass : 0;
-	MUTEX_EXIT(&ipf_frag);
-	return ret;
+	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;
 }
 
 
@@ -319,13 +369,32 @@ void *nat;
 	ipfr_t	*fr;
 	int	idx;
 
-	MUTEX_ENTER(&ipf_natfrag);
+	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;
 
-	MUTEX_EXIT(&ipf_natfrag);
+	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);
 }
 
 
@@ -334,55 +403,52 @@ void *nat;
  */
 void ipfr_unload()
 {
-	ipfr_t	**fp, *fr;
+	ipfr_t	**fp, *fra;
 	nat_t	*nat;
 	int	idx;
 
-	MUTEX_ENTER(&ipf_frag);
+	WRITE_ENTER(&ipf_frag);
 	for (idx = IPFT_SIZE - 1; idx >= 0; idx--)
-		for (fp = &ipfr_heads[idx]; (fr = *fp); ) {
-			*fp = fr->ipfr_next;
-			KFREE(fr);
+		for (fp = &ipfr_heads[idx]; (fra = *fp); ) {
+			*fp = fra->ipfr_next;
+			ipfr_delete(fra);
 		}
-	MUTEX_EXIT(&ipf_frag);
+	RWLOCK_EXIT(&ipf_frag);
 
-	MUTEX_ENTER(&ipf_nat);
-	MUTEX_ENTER(&ipf_natfrag);
+	WRITE_ENTER(&ipf_nat);
+	WRITE_ENTER(&ipf_natfrag);
 	for (idx = IPFT_SIZE - 1; idx >= 0; idx--)
-		for (fp = &ipfr_nattab[idx]; (fr = *fp); ) {
-			*fp = fr->ipfr_next;
-			if ((nat = (nat_t *)fr->ipfr_data)) {
-				if (nat->nat_data == fr)
+		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;
 			}
-			KFREE(fr);
+			ipfr_delete(fra);
 		}
-	MUTEX_EXIT(&ipf_natfrag);
-	MUTEX_EXIT(&ipf_nat);
+	RWLOCK_EXIT(&ipf_natfrag);
+	RWLOCK_EXIT(&ipf_nat);
 }
 
 
 #ifdef	_KERNEL
-/*
- * Slowly expire held state for fragments.  Timeouts are set * in expectation
- * of this being called twice per second.
- */
-# if (BSD >= 199306) || SOLARIS || defined(__sgi)
-void ipfr_slowtimer()
-# else
-int ipfr_slowtimer()
-# endif
+void ipfr_fragexpire()
 {
-	ipfr_t	**fp, *fr;
+	ipfr_t	**fp, *fra;
 	nat_t	*nat;
-	int	s, idx;
-
-#ifdef __sgi
-	ipfilter_sgi_intfsync();
+	int	idx;
+#if defined(_KERNEL)
+# if !SOLARIS
+	int	s;
+# endif
 #endif
 
+	if (fr_frag_lock)
+		return;
+
 	SPL_NET(s);
-	MUTEX_ENTER(&ipf_frag);
+	WRITE_ENTER(&ipf_frag);
 
 	/*
 	 * Go through the entire table, looking for entries to expire,
@@ -390,23 +456,17 @@ int ipfr_slowtimer()
 	 * remove it from the chain and free it.
 	 */
 	for (idx = IPFT_SIZE - 1; idx >= 0; idx--)
-		for (fp = &ipfr_heads[idx]; (fr = *fp); ) {
-			--fr->ipfr_ttl;
-			if (fr->ipfr_ttl == 0) {
-				if (fr->ipfr_prev)
-					fr->ipfr_prev->ipfr_next =
-					     fr->ipfr_next;
-				if (fr->ipfr_next)
-					fr->ipfr_next->ipfr_prev =
-					     fr->ipfr_prev;
-				*fp = fr->ipfr_next;
-				ipfr_stats.ifs_expire++;
-				ipfr_inuse--;
-				KFREE(fr);
+		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 = &fr->ipfr_next;
+				fp = &fra->ipfr_next;
 		}
-	MUTEX_EXIT(&ipf_frag);
+	RWLOCK_EXIT(&ipf_frag);
 
 	/*
 	 * Same again for the NAT table, except that if the structure also
@@ -415,44 +475,76 @@ int ipfr_slowtimer()
 	 * 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.
 	 */
-	MUTEX_ENTER(&ipf_nat);
-	MUTEX_ENTER(&ipf_natfrag);
+	WRITE_ENTER(&ipf_nat);
+	WRITE_ENTER(&ipf_natfrag);
 	for (idx = IPFT_SIZE - 1; idx >= 0; idx--)
-		for (fp = &ipfr_nattab[idx]; (fr = *fp); ) {
-			--fr->ipfr_ttl;
-			if (fr->ipfr_ttl == 0) {
-				if (fr->ipfr_prev)
-					fr->ipfr_prev->ipfr_next =
-					     fr->ipfr_next;
-				if (fr->ipfr_next)
-					fr->ipfr_next->ipfr_prev =
-					     fr->ipfr_prev;
-				*fp = fr->ipfr_next;
-				ipfr_stats.ifs_expire++;
-				ipfr_inuse--;
-				if ((nat = (nat_t *)fr->ipfr_data)) {
-					if (nat->nat_data == fr)
+		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;
 				}
-				KFREE(fr);
+				*fp = fra->ipfr_next;
+				ipfr_delete(fra);
 			} else
-				fp = &fr->ipfr_next;
+				fp = &fra->ipfr_next;
 		}
-	MUTEX_EXIT(&ipf_natfrag);
-	MUTEX_EXIT(&ipf_nat);
+	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
+# if    SOLARIS
 	ipfr_timer_id = timeout(ipfr_slowtimer, NULL, drv_usectohz(500000));
+	RWLOCK_EXIT(&ipf_solaris);
 # else
-#  ifndef linux
-	ip_slowtimo();
-#  endif
-#  if (BSD < 199306) && !defined(__sgi)
+#  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
-# endif
+#   endif /* FreeBSD */
+#  endif /* NetBSD */
+# endif /* SOLARIS */
 }
 #endif /* defined(_KERNEL) */