PERFORCE change 38253 for review

Sam Leffler sam at FreeBSD.org
Thu Sep 18 18:13:51 PDT 2003


http://perforce.freebsd.org/chv.cgi?CH=38253

Change 38253 by sam at sam_ebb on 2003/09/18 18:13:30

	PFIL_HOOKS revision; derived from work by
	"Max Laier" <max at love2party.net>:
	
	o update API to match netbsd
	o add locking (similar to the MAC)
	o revise call sites to use new API
	o wack ipfilter code to suit
	o bump FreeBSD version to identify API change
	o cleanup some grottiness in pfil.h for ipfilter that I'm sure
	  I'll be sorry I did
	
	This stuff needs more testing and review-by/handoff-to Darren.

Affected files ...

.. //depot/projects/netperf/sys/contrib/ipfilter/netinet/ip_fil.c#2 edit
.. //depot/projects/netperf/sys/modules/ipfilter/Makefile#2 edit
.. //depot/projects/netperf/sys/net/bridge.c#9 edit
.. //depot/projects/netperf/sys/net/pfil.c#2 edit
.. //depot/projects/netperf/sys/net/pfil.h#2 edit
.. //depot/projects/netperf/sys/netinet/ip_input.c#5 edit
.. //depot/projects/netperf/sys/netinet/ip_output.c#5 edit
.. //depot/projects/netperf/sys/netinet/ip_var.h#4 edit
.. //depot/projects/netperf/sys/netinet6/ip6_forward.c#2 edit
.. //depot/projects/netperf/sys/netinet6/ip6_input.c#3 edit
.. //depot/projects/netperf/sys/netinet6/ip6_output.c#4 edit
.. //depot/projects/netperf/sys/netinet6/ip6_var.h#2 edit
.. //depot/projects/netperf/sys/sys/param.h#6 edit

Differences ...

==== //depot/projects/netperf/sys/contrib/ipfilter/netinet/ip_fil.c#2 (text+ko) ====

@@ -307,6 +307,61 @@
 }
 # endif
 #endif /* __NetBSD_Version >= 105110000 && _KERNEL */
+#if (__FreeBSD_version >= 501108)
+# include <net/pfil.h>
+
+static int
+fr_check_wrapper(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
+{
+	struct ip *ip = mtod(*mp, struct ip *);
+	int rv, hlen = ip->ip_hl << 2;
+
+	/*
+	 * If the packet is out-bound, we can't delay checksums
+	 * here.  For in-bound, the checksum has already been
+	 * validated.
+	 */
+	if (dir == PFIL_OUT) {
+		if ((*mp)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
+			in_delayed_cksum(*mp);
+			(*mp)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
+		}
+	}
+
+	/*
+	 * We get the packet with all fields in network byte
+	 * order.  We expect ip_len and ip_off to be in host
+	 * order.  We frob them, call the filter, then frob
+	 * them back.
+	 *
+	 * Note, we don't need to update the checksum, because
+	 * it has already been verified.
+	 */
+	ip->ip_len = ntohs(ip->ip_len);
+	ip->ip_off = ntohs(ip->ip_off);
+
+	rv = fr_check(ip, hlen, ifp, (dir == PFIL_OUT), mp);
+
+	if (rv == 0 && *mp != NULL) {
+		ip = mtod(*mp, struct ip *);
+		ip->ip_len = ntohs(ip->ip_len);
+		ip->ip_off = ntohs(ip->ip_off);
+	}
+
+	return (rv);
+}
+
+# ifdef USE_INET6
+#  include <netinet/ip6.h>
+
+static int
+fr_check_wrapper6(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
+{
+	return (fr_check(mtod(*mp, struct ip *), sizeof(struct ip6_hdr),
+	    ifp, (dir == PFIL_OUT), mp));
+}
+# endif
+#endif /* __FreeBSD_version >= 501108 */
 #ifdef	_KERNEL
 # if	defined(IPFILTER_LKM) && !defined(__sgi)
 int iplidentify(s)
@@ -348,7 +403,8 @@
   ((__NetBSD_Version__ >= 104200000) || (__FreeBSD_version >= 500011)))
 	int error = 0;
 # endif
-#if defined(__NetBSD_Version__) && (__NetBSD_Version__ >= 105110000)
+#if (defined(__NetBSD_Version__) && (__NetBSD_Version__ >= 105110000)) || \
+    (__FreeBSD_version >= 501108)
 	struct pfil_head *ph_inet;
 # ifdef USE_INET6
 	struct pfil_head *ph_inet6;
@@ -380,7 +436,7 @@
 
 # ifdef NETBSD_PF
 #  if (__NetBSD_Version__ >= 104200000) || (__FreeBSD_version >= 500011)
-#   if __NetBSD_Version__ >= 105110000
+#   if (__NetBSD_Version__ >= 105110000) || (__FreeBSD_version >= 501108)
 	ph_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
 #    ifdef USE_INET6
 	ph_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
@@ -416,7 +472,7 @@
 	pfil_add_hook((void *)fr_check, PFIL_IN|PFIL_OUT);
 #  endif
 #  ifdef USE_INET6
-#   if __NetBSD_Version__ >= 105110000
+#   if (__NetBSD_Version__ >= 105110000) || (__FreeBSD_version >= 501108)
 	if (ph_inet6 != NULL)
 		error = pfil_add_hook((void *)fr_check_wrapper6, NULL,
 				      PFIL_IN|PFIL_OUT, ph_inet6);
@@ -508,7 +564,7 @@
 #if defined(NETBSD_PF) && \
     ((__NetBSD_Version__ >= 104200000) || (__FreeBSD_version >= 500011))
 	int error = 0;
-# if __NetBSD_Version__ >= 105150000
+# if (__NetBSD_Version__ >= 105150000) || (__FreeBSD_version >= 501108)
         struct pfil_head *ph_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
 #  ifdef USE_INET6
         struct pfil_head *ph_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
@@ -552,7 +608,7 @@
 
 # ifdef NETBSD_PF
 #  if ((__NetBSD_Version__ >= 104200000) || (__FreeBSD_version >= 500011))
-#   if __NetBSD_Version__ >= 105110000
+#   if (__NetBSD_Version__ >= 105110000) || (__FreeBSD_version >= 501108)
 	if (ph_inet != NULL)
 		error = pfil_remove_hook((void *)fr_check_wrapper, NULL,
 					 PFIL_IN|PFIL_OUT, ph_inet);

==== //depot/projects/netperf/sys/modules/ipfilter/Makefile#2 (text+ko) ====

@@ -7,7 +7,7 @@
 	ip_log.c ip_fil.c fil.c
 
 CFLAGS+=  -I${.CURDIR}/../../contrib/ipfilter
-CFLAGS+= -DIPFILTER=1 -DIPFILTER_LKM -DIPFILTER_LOG
+CFLAGS+= -DIPFILTER=1 -DIPFILTER_LKM -DIPFILTER_LOG -DPFIL_HOOKS
 #
 # If you don't want log functionality remove -DIPFILTER_LOG
 #

==== //depot/projects/netperf/sys/net/bridge.c#9 (text+ko) ====

@@ -87,6 +87,7 @@
  *  - be very careful when bridging VLANs
  *  - loop detection is still not very robust.
  */
+#include "opt_pfil_hooks.h"
 
 #include <sys/param.h>
 #include <sys/mbuf.h>
@@ -98,7 +99,6 @@
 #include <sys/kernel.h>
 #include <sys/sysctl.h>
 
-#include <net/pfil.h>	/* for ipfilter */
 #include <net/if.h>
 #include <net/if_types.h>
 #include <net/if_var.h>
@@ -109,6 +109,11 @@
 #include <netinet/ip.h>
 #include <netinet/if_ether.h> /* for struct arpcom */
 
+#ifdef PFIL_HOOKS
+#include <net/pfil.h>
+#include <netinet/ip_var.h>
+#endif
+
 #include <net/route.h>
 #include <netinet/ip_fw.h>
 #include <netinet/ip_dummynet.h>
@@ -160,7 +165,6 @@
 
 
 extern struct protosw inetsw[];			/* from netinet/ip_input.c */
-extern u_char ip_protox[];			/* from netinet/ip_input.c */
 
 static int n_clusters;				/* number of clusters */
 static struct cluster_softc *clusters;
@@ -715,6 +719,7 @@
      * Lookup local addresses in case one matches.  We optimize
      * for the common case of two interfaces.
      */
+    KASSERT(c->ports != 0, ("lookup with no ports!"));
     switch (c->ports) {
 	int i;
     default:
@@ -914,10 +919,6 @@
     int shared = bdg_copy;		/* someone else is using the mbuf */
     struct ifnet *real_dst = dst;	/* real dst from ether_output */
     struct ip_fw_args args;
-#ifdef PFIL_HOOKS
-    struct packet_filter_hook *pfh;
-    int rv;
-#endif /* PFIL_HOOKS */
     struct ether_header save_eh;
     struct mbuf *m;
 
@@ -969,7 +970,7 @@
      */
     if (src != NULL && (
 #ifdef PFIL_HOOKS
-	((pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh)) != NULL && bdg_ipf !=0) ||
+	(inet_pfil_hook.ph_busy_count >= 0 && bdg_ipf != 0) ||
 #endif
 	(IPFW_LOADED && bdg_ipfw != 0))) {
 
@@ -1006,8 +1007,9 @@
 	 * NetBSD-style generic packet filter, pfil(9), hooks.
 	 * Enables ipf(8) in bridging.
 	 */
-	if (pfh != NULL && m0->m_pkthdr.len >= sizeof(struct ip) &&
-		ntohs(save_eh.ether_type) == ETHERTYPE_IP) {
+	if (inet_pfil_hook.ph_busy_count >= 0 &&
+	    m0->m_pkthdr.len >= sizeof(struct ip) &&
+	    ntohs(save_eh.ether_type) == ETHERTYPE_IP) {
 	    /*
 	     * before calling the firewall, swap fields the same as IP does.
 	     * here we assume the pkt is an IP one and the header is contiguous
@@ -1017,20 +1019,14 @@
 	    ip->ip_len = ntohs(ip->ip_len);
 	    ip->ip_off = ntohs(ip->ip_off);
 
-	    do {
-		if (pfh->pfil_func) {
-		    rv = pfh->pfil_func(ip, ip->ip_hl << 2, src, 0, &m0);
-		    if (m0 == NULL) {
-			bdg_dropped++;
-			return NULL;
-		    }
-		    if (rv != 0) {
-			EH_RESTORE(m0);		/* restore Ethernet header */
-			return m0;
-		    }
-		    ip = mtod(m0, struct ip *);
-		}
-	    } while ((pfh = TAILQ_NEXT(pfh, pfil_link)) != NULL);
+	    if (pfil_run_hooks(&inet_pfil_hook, &m0, src, PFIL_IN) != 0) {
+		EH_RESTORE(m0);		/* restore Ethernet header */
+		return m0;
+	    }
+	    if (m0 == NULL) {
+		bdg_dropped++;
+		return NULL;
+	    }
 	    /*
 	     * If we get here, the firewall has passed the pkt, but the mbuf
 	     * pointer might have changed. Restore ip and the fields ntohs()'d.

==== //depot/projects/netperf/sys/net/pfil.c#2 (text+ko) ====

@@ -1,4 +1,5 @@
-/*	$FreeBSD: src/sys/net/pfil.c,v 1.7 2003/02/19 05:47:29 imp Exp $	*/
+/*	$FreeBSD$ */
+/*	$NetBSD: pfil.c,v 1.20 2001/11/12 23:49:46 lukem Exp $	*/
 
 /*
  * Copyright (c) 1996 Matthew R. Green
@@ -29,30 +30,207 @@
  */
 
 #include <sys/param.h>
+#include <sys/kernel.h>
 #include <sys/errno.h>
 #include <sys/malloc.h>
 #include <sys/socket.h>
 #include <sys/socketvar.h>
 #include <sys/systm.h>
+#include <sys/condvar.h>
+#include <sys/lock.h>
+#include <sys/mutex.h>
+#include <sys/proc.h>
 #include <sys/queue.h>
 
 #include <net/if.h>
 #include <net/pfil.h>
 
-static void pfil_init(struct pfil_head *);
-static int pfil_list_add(pfil_list_t *,
-    int (*)(void *, int, struct ifnet *, int, struct mbuf **), int);
+static struct mtx pfil_global_lock;
+
+MTX_SYSINIT(pfil_heads_lock, &pfil_global_lock, "pfil_head_list lock", MTX_DEF);
+
+static int pfil_list_add(pfil_list_t *, struct packet_filter_hook *, int);
+
 static int pfil_list_remove(pfil_list_t *,
-    int (*)(void *, int, struct ifnet *, int, struct mbuf **));
+    int (*)(void *, struct mbuf **, struct ifnet *, int), void *);
+
+LIST_HEAD(, pfil_head) pfil_head_list =
+    LIST_HEAD_INITIALIZER(&pfil_head_list);
+
+static __inline void
+PFIL_RLOCK(struct pfil_head *ph)
+{
+	mtx_lock(&ph->ph_mtx);
+	ph->ph_busy_count++;
+	mtx_unlock(&ph->ph_mtx);
+}
+
+static __inline void
+PFIL_RUNLOCK(struct pfil_head *ph)
+{
+	mtx_lock(&ph->ph_mtx);
+	ph->ph_busy_count--;
+	if (ph->ph_busy_count == 0 && ph->ph_want_write)
+		cv_signal(&ph->ph_cv);
+	mtx_unlock(&ph->ph_mtx);
+}
+
+static __inline void
+PFIL_WLOCK(struct pfil_head *ph)
+{
+	mtx_lock(&ph->ph_mtx);
+	ph->ph_want_write = 1;
+	while (ph->ph_busy_count > 0)
+		cv_wait(&ph->ph_cv, &ph->ph_mtx);
+}
+
+static __inline int
+PFIL_TRY_WLOCK(struct pfil_head *ph)
+{
+	mtx_lock(&ph->ph_mtx);
+	ph->ph_want_write = 1;
+	if (ph->ph_busy_count > 0) {
+		ph->ph_want_write = 0;
+		mtx_unlock(&ph->ph_mtx);
+		return EBUSY;
+	}
+	return 0;
+}
+
+static __inline void
+PFIL_WUNLOCK(struct pfil_head *ph)
+{
+	ph->ph_want_write = 0;				\
+	mtx_unlock(&ph->ph_mtx);
+	cv_signal(&ph->ph_cv);
+}
+
+#define PFIL_LIST_LOCK() mtx_lock(&pfil_global_lock)
+#define PFIL_LIST_UNLOCK() mtx_unlock(&pfil_global_lock)
+
+/*
+ * pfil_run_hooks() runs the specified packet filter hooks.
+ */
+int
+pfil_run_hooks(struct pfil_head *ph, struct mbuf **mp, struct ifnet *ifp,
+    int dir)
+{
+	struct packet_filter_hook *pfh;
+	struct mbuf *m = *mp;
+	int rv = 0, lockgiant;
+
+	if (ph->ph_busy_count == -1 || ph->ph_want_write)
+		return (0);
+
+	lockgiant = dir & PFIL_GIANT;
+	if (lockgiant) {
+		mtx_lock(&Giant);
+		dir &= ~PFIL_GIANT;
+	}
+	
+	PFIL_RLOCK(ph);
+	for (pfh = pfil_hook_get(dir, ph); pfh != NULL;
+	     pfh = TAILQ_NEXT(pfh, pfil_link)) {
+		if (pfh->pfil_func != NULL) {
+			rv = (*pfh->pfil_func)(pfh->pfil_arg, &m, ifp, dir);
+			if (rv != 0 || m == NULL)
+				break;
+		}
+	}
+	PFIL_RUNLOCK(ph);
+
+	if (lockgiant)
+		mtx_unlock(&Giant);
+	
+	*mp = m;
+	return (rv);
+}
 
-static void
-pfil_init(ph)
-	 struct pfil_head *ph;
+/*
+ * pfil_head_register() registers a pfil_head with the packet filter
+ * hook mechanism.
+ */
+int
+pfil_head_register(struct pfil_head *ph)
 {
+	struct pfil_head *lph;
 
+	PFIL_LIST_LOCK();
+	LIST_FOREACH(lph, &pfil_head_list, ph_list)
+		if (ph->ph_type == lph->ph_type &&
+		    ph->ph_un.phu_val == lph->ph_un.phu_val) {
+			PFIL_LIST_UNLOCK();
+			return EEXIST;
+		}
+	PFIL_LIST_UNLOCK();
+
+	if (mtx_initialized(&ph->ph_mtx)) {	/* should not happen */
+		KASSERT((0), ("%s: allready initialized!", __func__));
+		return EBUSY;
+	} else {
+		ph->ph_busy_count = -1;
+		ph->ph_want_write = 1;
+		mtx_init(&ph->ph_mtx, "pfil_head_mtx", NULL, MTX_DEF);
+		cv_init(&ph->ph_cv, "pfil_head_cv");
+		mtx_lock(&ph->ph_mtx);			/* XXX: race? */
+	}
+
 	TAILQ_INIT(&ph->ph_in);
 	TAILQ_INIT(&ph->ph_out);
-	ph->ph_init = 1;
+
+	PFIL_LIST_LOCK();
+	LIST_INSERT_HEAD(&pfil_head_list, ph, ph_list);
+	PFIL_LIST_UNLOCK();
+	
+	PFIL_WUNLOCK(ph);
+	
+	return (0);
+}
+
+/*
+ * pfil_head_unregister() removes a pfil_head from the packet filter
+ * hook mechanism.
+ */
+int
+pfil_head_unregister(struct pfil_head *ph)
+{
+	struct packet_filter_hook *pfh, *pfnext;
+		
+	PFIL_LIST_LOCK();
+	/* 
+	 * LIST_REMOVE is safe for unlocked pfil_heads in ph_list.
+	 * No need to WLOCK all of them.
+	 */
+	LIST_REMOVE(ph, ph_list);
+	PFIL_LIST_UNLOCK();
+
+	PFIL_WLOCK(ph);			/* XXX: may sleep (cv_wait)! */
+	
+	TAILQ_FOREACH_SAFE(pfh, &ph->ph_in, pfil_link, pfnext)
+		free(pfh, M_IFADDR);
+	TAILQ_FOREACH_SAFE(pfh, &ph->ph_out, pfil_link, pfnext)
+		free(pfh, M_IFADDR);
+	cv_destroy(&ph->ph_cv);
+	mtx_destroy(&ph->ph_mtx);
+	
+	return (0);
+}
+
+/*
+ * pfil_head_get() returns the pfil_head for a given key/dlt.
+ */
+struct pfil_head *
+pfil_head_get(int type, u_long val)
+{
+	struct pfil_head *ph;
+
+	PFIL_LIST_LOCK();
+	LIST_FOREACH(ph, &pfil_head_list, ph_list)
+		if (ph->ph_type == type && ph->ph_un.phu_val == val)
+			break;
+	PFIL_LIST_UNLOCK();
+	
+	return (ph);
 }
 
 /*
@@ -64,53 +242,71 @@
  *	PFIL_WAITOK	OK to call malloc with M_WAITOK.
  */
 int
-pfil_add_hook(func, flags, ph)
-	int	(*func)(void *, int, struct ifnet *, int, struct mbuf **);
-	int	flags;
-	struct	pfil_head	*ph;
+pfil_add_hook(int (*func)(void *, struct mbuf **, struct ifnet *, int),
+    void *arg, int flags, struct pfil_head *ph)
 {
-	int err = 0;
+	struct packet_filter_hook *pfh1 = NULL;
+	struct packet_filter_hook *pfh2 = NULL;
+	int err;
+
+	/* Get memory */
+	if (flags & PFIL_IN) {
+		pfh1 = (struct packet_filter_hook *)malloc(sizeof(*pfh1), 
+		    M_IFADDR, (flags & PFIL_WAITOK) ? M_WAITOK : M_NOWAIT);
+		if (pfh1 == NULL) {
+			err = ENOMEM;
+			goto error;
+		}
+	}
+	if (flags & PFIL_OUT) {
+		pfh2 = (struct packet_filter_hook *)malloc(sizeof(*pfh1),
+		    M_IFADDR, (flags & PFIL_WAITOK) ? M_WAITOK : M_NOWAIT);
+		if (pfh2 == NULL) {
+			err = ENOMEM;
+			goto error;
+		}
+	}
 
-	if (ph->ph_init == 0)
-		pfil_init(ph);
+	/* Lock */
+	if (flags & PFIL_WAITOK)
+		PFIL_WLOCK(ph);
+	else {
+		err = PFIL_TRY_WLOCK(ph);
+		if (err)
+			goto error;
+	}
 
-	if (flags & PFIL_IN)
-		err = pfil_list_add(&ph->ph_in, func, flags & ~PFIL_OUT);
-	if (err)
-		return err;
-	if (flags & PFIL_OUT)
-		err = pfil_list_add(&ph->ph_out, func, flags & ~PFIL_IN);
-	if (err) {
-		if (flags & PFIL_IN)
-			pfil_list_remove(&ph->ph_in, func);
-		return err;
+	/* Add */
+	if (flags & PFIL_IN) {
+		pfh1->pfil_func = func;
+		pfh1->pfil_arg = arg;
+		err = pfil_list_add(&ph->ph_in, pfh1, flags & ~PFIL_OUT);
+		if (err)
+			goto done;
+	}
+	if (flags & PFIL_OUT) {
+		pfh2->pfil_func = func;
+		pfh2->pfil_arg = arg;
+		err = pfil_list_add(&ph->ph_out, pfh2, flags & ~PFIL_IN);
+		if (err) {
+			if (flags & PFIL_IN)
+				pfil_list_remove(&ph->ph_in, func, arg);
+			goto done;
+		}
 	}
-	return 0;
-}
 
-static int
-pfil_list_add(list, func, flags)
-	pfil_list_t *list;
-	int	(*func)(void *, int, struct ifnet *, int, struct mbuf **);
-	int flags;
-{
-	struct packet_filter_hook *pfh;
+	ph->ph_busy_count = 0;
+	PFIL_WUNLOCK(ph);
 
-	pfh = (struct packet_filter_hook *)malloc(sizeof(*pfh), M_IFADDR,
-	    flags & PFIL_WAITOK ? M_WAITOK : M_NOWAIT);
-	if (pfh == NULL)
-		return ENOMEM;
-	pfh->pfil_func = func;
-	/*
-	 * insert the input list in reverse order of the output list
-	 * so that the same path is followed in or out of the kernel.
-	 */
-	
-	if (flags & PFIL_IN)
-		TAILQ_INSERT_HEAD(list, pfh, pfil_link);
-	else
-		TAILQ_INSERT_TAIL(list, pfh, pfil_link);
 	return 0;
+done:
+	PFIL_WUNLOCK(ph);
+error:
+	if (pfh1 != NULL)
+		free(pfh1, M_IFADDR);
+	if (pfh2 != NULL)
+		free(pfh2, M_IFADDR);
+	return err;
 }
 
 /*
@@ -118,54 +314,71 @@
  * hook list.
  */
 int
-pfil_remove_hook(func, flags, ph)
-	int	(*func)(void *, int, struct ifnet *, int, struct mbuf **);
-	int	flags;
-	struct	pfil_head	*ph;
+pfil_remove_hook(int (*func)(void *, struct mbuf **, struct ifnet *, int),
+    void *arg, int flags, struct pfil_head *ph)
 {
 	int err = 0;
 
-	if (ph->ph_init == 0)
-		pfil_init(ph);
+	if (flags & PFIL_WAITOK)
+		PFIL_WLOCK(ph);
+	else {
+		err = PFIL_TRY_WLOCK(ph);
+		if (err)
+			return err;
+	}
 
 	if (flags & PFIL_IN)
-		err = pfil_list_remove(&ph->ph_in, func);
+		err = pfil_list_remove(&ph->ph_in, func, arg);
 	if ((err == 0) && (flags & PFIL_OUT))
-		err = pfil_list_remove(&ph->ph_out, func);
+		err = pfil_list_remove(&ph->ph_out, func, arg);
+
+	if (TAILQ_EMPTY(&ph->ph_in) && TAILQ_EMPTY(&ph->ph_out))
+		ph->ph_busy_count = -1;
+
+	PFIL_WUNLOCK(ph);
+	
 	return err;
 }
 
+static int
+pfil_list_add(pfil_list_t *list, struct packet_filter_hook *pfh1, int flags)
+{
+	struct packet_filter_hook *pfh;
+
+	/*
+	 * First make sure the hook is not already there.
+	 */
+	TAILQ_FOREACH(pfh, list, pfil_link)
+		if (pfh->pfil_func == pfh1->pfil_func &&
+		    pfh->pfil_arg == pfh1->pfil_arg)
+			return EEXIST;
+	/*
+	 * insert the input list in reverse order of the output list
+	 * so that the same path is followed in or out of the kernel.
+	 */
+	if (flags & PFIL_IN)
+		TAILQ_INSERT_HEAD(list, pfh1, pfil_link);
+	else
+		TAILQ_INSERT_TAIL(list, pfh1, pfil_link);
+
+	return 0;
+}
+
 /*
  * pfil_list_remove is an internal function that takes a function off the
  * specified list.
  */
 static int
-pfil_list_remove(list, func)
-	pfil_list_t *list;
-	int	(*func)(void *, int, struct ifnet *, int, struct mbuf **);
+pfil_list_remove(pfil_list_t *list,
+    int (*func)(void *, struct mbuf **, struct ifnet *, int), void *arg)
 {
 	struct packet_filter_hook *pfh;
 
 	TAILQ_FOREACH(pfh, list, pfil_link)
-		if (pfh->pfil_func == func) {
+		if (pfh->pfil_func == func && pfh->pfil_arg == arg) {
 			TAILQ_REMOVE(list, pfh, pfil_link);
 			free(pfh, M_IFADDR);
 			return 0;
 		}
 	return ENOENT;
 }
-
-struct packet_filter_hook *
-pfil_hook_get(flag, ph)
-	int flag;
-	struct	pfil_head	*ph;
-{
-	if (ph->ph_init != 0)
-		switch (flag) {
-		case PFIL_IN:
-			return (TAILQ_FIRST(&ph->ph_in));
-		case PFIL_OUT:
-			return (TAILQ_FIRST(&ph->ph_out));
-		}
-	return NULL;
-}

==== //depot/projects/netperf/sys/net/pfil.h#2 (text+ko) ====

@@ -1,4 +1,5 @@
-/*	$FreeBSD: src/sys/net/pfil.h,v 1.9 2002/03/24 09:34:04 bde Exp $	*/
+/*	$FreeBSD$ */
+/*	$NetBSD: pfil.h,v 1.22 2003/06/23 12:57:08 martin Exp $	*/
 
 /*
  * Copyright (c) 1996 Matthew R. Green
@@ -31,6 +32,10 @@
 #ifndef _NET_PFIL_H_
 #define _NET_PFIL_H_
 
+#include <sys/systm.h>
+#include <sys/_lock.h>
+#include <sys/_mutex.h>
+#include <sys/condvar.h>	/* XXX */
 #include <sys/queue.h>
 
 struct mbuf;
@@ -42,39 +47,68 @@
  */
 struct packet_filter_hook {
         TAILQ_ENTRY(packet_filter_hook) pfil_link;
-	int	(*pfil_func)(void *, int, struct ifnet *, int, struct mbuf **);
+	int	(*pfil_func)(void *, struct mbuf **, struct ifnet *, int);
+	void	*pfil_arg;
 	int	pfil_flags;
 };
 
 #define PFIL_IN		0x00000001
 #define PFIL_OUT	0x00000002
 #define PFIL_WAITOK	0x00000004
+/* XXX temporary until Giant gets removed */
+#define	PFIL_GIANT	0x00000008	/* lock Giant around calls */
 #define PFIL_ALL	(PFIL_IN|PFIL_OUT)
 
 typedef	TAILQ_HEAD(pfil_list, packet_filter_hook) pfil_list_t;
 
+#define	PFIL_TYPE_AF		1	/* key is AF_* type */
+#define	PFIL_TYPE_IFNET		2	/* key is ifnet pointer */
+
 struct pfil_head {
 	pfil_list_t	ph_in;
 	pfil_list_t	ph_out;
-	int		ph_init;
+	int		ph_type;
+	/*
+	 * Locking: use a busycounter per pfil_head.
+	 * Use ph_busy_count = -1 to indicate pfil_head is empty.
+	 */
+	int		ph_busy_count;	/* count of threads with read lock */
+	int		ph_want_write;	/* want write lock flag */
+	struct cv	ph_cv;		/* for waking up writers */
+	struct mtx	ph_mtx;		/* mutex on locking state */
+	union {
+		u_long		phu_val;
+		void		*phu_ptr;
+	} ph_un;
+#define	ph_af		ph_un.phu_val
+#define	ph_ifnet	ph_un.phu_ptr
+	LIST_ENTRY(pfil_head) ph_list;
 };
 
-struct packet_filter_hook *pfil_hook_get(int, struct pfil_head *);
-int	pfil_add_hook(int (*func)(void *, int,
-	    struct ifnet *, int, struct mbuf **), int, struct pfil_head *);
-int	pfil_remove_hook(int (*func)(void *, int,
-	    struct ifnet *, int, struct mbuf **), int, struct pfil_head *);
+int	pfil_run_hooks(struct pfil_head *, struct mbuf **, struct ifnet *,
+	    int);
+
+int	pfil_add_hook(int (*func)(void *, struct mbuf **,
+	    struct ifnet *, int), void *, int, struct pfil_head *);
+int	pfil_remove_hook(int (*func)(void *, struct mbuf **,
+	    struct ifnet *, int), void *, int, struct pfil_head *);
+
+int	pfil_head_register(struct pfil_head *);
+int	pfil_head_unregister(struct pfil_head *);
 
-/* XXX */
-#if defined(_KERNEL) && !defined(KLD_MODULE)
-#include "opt_ipfilter.h"
-#endif
+struct pfil_head *pfil_head_get(int, u_long);
 
-#if IPFILTER > 0
-#ifdef PFIL_HOOKS
-#undef PFIL_HOOKS
-#endif
-#define PFIL_HOOKS
-#endif /* IPFILTER */
+static __inline struct packet_filter_hook *
+pfil_hook_get(int dir, struct pfil_head *ph)
+{
+	KASSERT(ph->ph_busy_count > 0, 
+	    ("pfil_hook_get: called on unbusy pfil_head"));
+	if (dir == PFIL_IN)
+		return (TAILQ_FIRST(&ph->ph_in));
+	else if (dir == PFIL_OUT)
+		return (TAILQ_FIRST(&ph->ph_out));
+	else
+		return (NULL);
+}
 
 #endif /* _NET_PFIL_H_ */

==== //depot/projects/netperf/sys/netinet/ip_input.c#5 (text+ko) ====

@@ -154,6 +154,9 @@
 #ifdef DIAGNOSTIC
 static int	ipprintfs = 0;
 #endif
+#ifdef PFIL_HOOKS
+struct pfil_head inet_pfil_hook;
+#endif
 
 static struct	ifqueue ipintrq;
 static int	ipqmaxlen = IFQ_MAXLEN;
@@ -263,6 +266,14 @@
 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
 			ip_protox[pr->pr_protocol] = pr - inetsw;
 
+#ifdef PFIL_HOOKS
+	inet_pfil_hook.ph_type = PFIL_TYPE_AF;
+	inet_pfil_hook.ph_af = AF_INET;
+	if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
+		printf("%s: WARNING: unable to register pfil hook, "
+			"error %d\n", __func__, i);
+#endif /* PFIL_HOOKS */
+
 	IPQ_LOCK_INIT();
 	for (i = 0; i < IPREASS_NHASH; i++)
 	    TAILQ_INIT(&ipq[i]);
@@ -301,11 +312,6 @@
 	struct in_addr pkt_dst;
 	u_int32_t divert_info = 0;		/* packet divert/tee info */
 	struct ip_fw_args args;
-#ifdef PFIL_HOOKS
-	struct packet_filter_hook *pfh;
-	struct mbuf *m0;
-	int rv;
-#endif /* PFIL_HOOKS */
 #ifdef FAST_IPSEC
 	struct m_tag *mtag;
 	struct tdb_ident *tdbi;
@@ -463,28 +469,14 @@
 
 #ifdef PFIL_HOOKS
 	/*
-	 * Run through list of hooks for input packets.  If there are any
-	 * filters which require that additional packets in the flow are
-	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
-	 * Note that filters must _never_ set this flag, as another filter
-	 * in the list may have previously cleared it.
+	 * Run through list of hooks for input packets.
 	 */
-	m0 = m;
-	/* XXX locking */
-	pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
-	for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link))
-		if (pfh->pfil_func) {
-			mtx_lock(&Giant);	/* XXX */
-			rv = pfh->pfil_func(ip, hlen,
-					    m->m_pkthdr.rcvif, 0, &m0);
-			mtx_unlock(&Giant);	/* XXX */
-			if (rv)
-				return;
-			m = m0;
-			if (m == NULL)
-				return;
-			ip = mtod(m, struct ip *);
-		}
+	if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
+	    PFIL_IN | PFIL_GIANT) != 0)
+		return;
+	if (m == NULL)			/* consumed by filter */
+		return;
+	ip = mtod(m, struct ip *);
 #endif /* PFIL_HOOKS */
 
 	if (fw_enable && IPFW_LOADED) {

==== //depot/projects/netperf/sys/netinet/ip_output.c#5 (text+ko) ====

@@ -149,11 +149,6 @@
 #endif /* FAST_IPSEC */
 	struct ip_fw_args args;
 	int src_was_INADDR_ANY = 0;	/* as the name says... */
-#ifdef PFIL_HOOKS
-	struct packet_filter_hook *pfh;
-	struct mbuf *m1;
-	int rv;
-#endif /* PFIL_HOOKS */
 
 	args.eh = NULL;
 	args.rule = NULL;
@@ -741,20 +736,10 @@
 	/*
 	 * Run through list of hooks for output packets.
 	 */
-	m1 = m;
-	pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
-	for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link))
-		if (pfh->pfil_func) {
-			rv = pfh->pfil_func(ip, hlen, ifp, 1, &m1);
-			if (rv) {
-				error = EHOSTUNREACH;
-				goto done;
-			}
-			m = m1;
-			if (m == NULL)
-				goto done;
-			ip = mtod(m, struct ip *);
-		}
+	error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT);
+	if (error != 0 || m == NULL)
+		goto done;
+	ip = mtod(m, struct ip *);
 #endif /* PFIL_HOOKS */
 
 	/*

==== //depot/projects/netperf/sys/netinet/ip_var.h#4 (text+ko) ====

@@ -207,6 +207,10 @@
 extern struct pr_usrreqs div_usrreqs;
 #endif
 
+#ifdef PFIL_HOOKS
+extern	struct pfil_head inet_pfil_hook;
+#endif
+
 void	in_delayed_cksum(struct mbuf *m);
 
 #endif /* _KERNEL */

==== //depot/projects/netperf/sys/netinet6/ip6_forward.c#2 (text+ko) ====

@@ -34,6 +34,7 @@
 #include "opt_inet.h"
 #include "opt_inet6.h"
 #include "opt_ipsec.h"
+#include "opt_pfil_hooks.h"
 
 #include <sys/param.h>
 #include <sys/systm.h>
@@ -113,11 +114,6 @@
 	int error, type = 0, code = 0;
 	struct mbuf *mcopy = NULL;
 	struct ifnet *origifp;	/* maybe unnecessary */
-#ifdef PFIL_HOOKS
-	struct packet_filter_hook *pfh;
-	struct mbuf *m1;
-	int rv;
-#endif /* PFIL_HOOKS */
 #ifdef IPSEC
 	struct secpolicy *sp = NULL;
 #endif
@@ -526,21 +522,13 @@
 	/*
 	 * Run through list of hooks for output packets.
 	 */
-	m1 = m;
-	pfh = pfil_hook_get(PFIL_OUT, &inet6sw[ip6_protox[IPPROTO_IPV6]].pr_pfh);
-	for (; pfh; pfh = pfh->pfil_link.tqe_next)
-		if (pfh->pfil_func) {
-			rv = pfh->pfil_func(ip6, sizeof(*ip6),
-					    rt->rt_ifp, 1, &m1);
-			if (rv) {
-				error = EHOSTUNREACH;
-				goto freecopy;
-			}
-			m = m1;
-			if (m == NULL)
-				goto freecopy;
-			ip6 = mtod(m, struct ip6_hdr *);
-		}
+	if (pfil_run_hooks(&inet6_pfil_hook, &m, rt->rt_ifp, PFIL_OUT) != 0) {
+		error = EHOSTUNREACH;
+		goto freecopy;
+	}
+	if (m == NULL)
+		goto freecopy;
+	ip6 = mtod(m, struct ip6_hdr *);
 #endif /* PFIL_HOOKS */
 
 	error = nd6_output(rt->rt_ifp, origifp, m, dst, rt);

==== //depot/projects/netperf/sys/netinet6/ip6_input.c#3 (text+ko) ====

@@ -143,6 +143,9 @@
 
 int ip6_ours_check_algorithm;
 
+#ifdef PFIL_HOOKS
+struct pfil_head inet6_pfil_hook;
+#endif
 

>>> TRUNCATED FOR MAIL (1000 lines) <<<


More information about the p4-projects mailing list