Overhaul if_gif(4):

 o convert to if_transmit;
 o use rmlock to protect access to gif_softc;
 o use sx lock to protect from concurrent ioctls;
 o remove a lot of unneeded and duplicated code;
 o remove cached route support (it won't work with concurrent io);
 o style fixes.

Reviewed by:	melifaro
Obtained from:	Yandex LLC
MFC after:	1 month
Sponsored by:	Yandex LLC

1 files changed, 45 insertions, 264 deletions

diff --git a/sys/netinet6/in6_gif.c b/sys/netinet6/in6_gif.c
index b792ed5..6d48b5d 100644
--- a/sys/netinet6/in6_gif.c
+++ b/sys/netinet6/in6_gif.c
@@ -36,6 +36,8 @@ __FBSDID("$FreeBSD$");
 #include "opt_inet6.h"
 
 #include <sys/param.h>
+#include <sys/lock.h>
+#include <sys/rmlock.h>
 #include <sys/systm.h>
 #include <sys/socket.h>
 #include <sys/sockio.h>
@@ -95,113 +97,26 @@ struct protosw in6_gif_protosw = {
 };
 
 int
-in6_gif_output(struct ifnet *ifp,
-    int family,			/* family of the packet to be encapsulate */
-    struct mbuf *m)
+in6_gif_output(struct ifnet *ifp, struct mbuf *m, int proto, uint8_t ecn)
 {
+	GIF_RLOCK_TRACKER;
 	struct gif_softc *sc = ifp->if_softc;
-	struct sockaddr_in6 *dst = (struct sockaddr_in6 *)&sc->gif_ro6.ro_dst;
-	struct sockaddr_in6 *sin6_src = (struct sockaddr_in6 *)sc->gif_psrc;
-	struct sockaddr_in6 *sin6_dst = (struct sockaddr_in6 *)sc->gif_pdst;
 	struct ip6_hdr *ip6;
-	struct etherip_header eiphdr;
-	int error, len, proto;
-	u_int8_t itos, otos;
-
-	GIF_LOCK_ASSERT(sc);
-
-	if (sin6_src == NULL || sin6_dst == NULL ||
-	    sin6_src->sin6_family != AF_INET6 ||
-	    sin6_dst->sin6_family != AF_INET6) {
-		m_freem(m);
-		return EAFNOSUPPORT;
-	}
-
-	switch (family) {
-#ifdef INET
-	case AF_INET:
-	    {
-		struct ip *ip;
-
-		proto = IPPROTO_IPV4;
-		if (m->m_len < sizeof(*ip)) {
-			m = m_pullup(m, sizeof(*ip));
-			if (!m)
-				return ENOBUFS;
-		}
-		ip = mtod(m, struct ip *);
-		itos = ip->ip_tos;
-		break;
-	    }
-#endif
-#ifdef INET6
-	case AF_INET6:
-	    {
-		struct ip6_hdr *ip6;
-		proto = IPPROTO_IPV6;
-		if (m->m_len < sizeof(*ip6)) {
-			m = m_pullup(m, sizeof(*ip6));
-			if (!m)
-				return ENOBUFS;
-		}
-		ip6 = mtod(m, struct ip6_hdr *);
-		itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
-		break;
-	    }
-#endif
-	case AF_LINK:
-		proto = IPPROTO_ETHERIP;
-
-		/*
-		 * GIF_SEND_REVETHIP (disabled by default) intentionally
-		 * sends an EtherIP packet with revered version field in
-		 * the header.  This is a knob for backward compatibility
-		 * with FreeBSD 7.2R or prior.
-		 */
-		if ((sc->gif_options & GIF_SEND_REVETHIP)) {
-			eiphdr.eip_ver = 0;
-			eiphdr.eip_resvl = ETHERIP_VERSION;
-			eiphdr.eip_resvh = 0;
-		} else {
-			eiphdr.eip_ver = ETHERIP_VERSION;
-			eiphdr.eip_resvl = 0;
-			eiphdr.eip_resvh = 0;
-		}
-		/* prepend Ethernet-in-IP header */
-		M_PREPEND(m, sizeof(struct etherip_header), M_NOWAIT);
-		if (m && m->m_len < sizeof(struct etherip_header))
-			m = m_pullup(m, sizeof(struct etherip_header));
-		if (m == NULL)
-			return ENOBUFS;
-		bcopy(&eiphdr, mtod(m, struct etherip_header *),
-		    sizeof(struct etherip_header));
-		itos = 0;
-		break;
-
-	default:
-#ifdef DEBUG
-		printf("in6_gif_output: warning: unknown family %d passed\n",
-			family);
-#endif
-		m_freem(m);
-		return EAFNOSUPPORT;
-	}
+	int len;
 
 	/* prepend new IP header */
 	len = sizeof(struct ip6_hdr);
 #ifndef __NO_STRICT_ALIGNMENT
-	if (family == AF_LINK)
+	if (proto == IPPROTO_ETHERIP)
 		len += ETHERIP_ALIGN;
 #endif
 	M_PREPEND(m, len, M_NOWAIT);
 	if (m != NULL && m->m_len < len)
 		m = m_pullup(m, len);
-	if (m == NULL) {
-		printf("ENOBUFS in in6_gif_output %d\n", __LINE__);
-		return ENOBUFS;
-	}
+	if (m == NULL)
+		return (ENOBUFS);
 #ifndef __NO_STRICT_ALIGNMENT
-	if (family == AF_LINK) {
+	if (proto == IPPROTO_ETHERIP) {
 		len = mtod(m, vm_offset_t) & 3;
 		KASSERT(len == 0 || len == ETHERIP_ALIGN,
 		    ("in6_gif_output: unexpected misalignment"));
@@ -211,166 +126,52 @@ in6_gif_output(struct ifnet *ifp,
 #endif
 
 	ip6 = mtod(m, struct ip6_hdr *);
-	ip6->ip6_flow	= 0;
-	ip6->ip6_vfc	&= ~IPV6_VERSION_MASK;
-	ip6->ip6_vfc	|= IPV6_VERSION;
-	ip6->ip6_plen	= htons((u_short)m->m_pkthdr.len);
-	ip6->ip6_nxt	= proto;
-	ip6->ip6_hlim	= V_ip6_gif_hlim;
-	ip6->ip6_src	= sin6_src->sin6_addr;
-	/* bidirectional configured tunnel mode */
-	if (!IN6_IS_ADDR_UNSPECIFIED(&sin6_dst->sin6_addr))
-		ip6->ip6_dst = sin6_dst->sin6_addr;
-	else  {
+	GIF_RLOCK(sc);
+	if (sc->gif_family != AF_INET6) {
 		m_freem(m);
-		return ENETUNREACH;
+		GIF_RUNLOCK(sc);
+		return (ENETDOWN);
 	}
-	ip_ecn_ingress((ifp->if_flags & IFF_LINK1) ? ECN_ALLOWED : ECN_NOCARE,
-		       &otos, &itos);
-	ip6->ip6_flow &= ~htonl(0xff << 20);
-	ip6->ip6_flow |= htonl((u_int32_t)otos << 20);
+	bcopy(sc->gif_ip6hdr, ip6, sizeof(struct ip6_hdr));
+	GIF_RUNLOCK(sc);
 
-	M_SETFIB(m, sc->gif_fibnum);
-
-	if (dst->sin6_family != sin6_dst->sin6_family ||
-	     !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &sin6_dst->sin6_addr)) {
-		/* cache route doesn't match */
-		bzero(dst, sizeof(*dst));
-		dst->sin6_family = sin6_dst->sin6_family;
-		dst->sin6_len = sizeof(struct sockaddr_in6);
-		dst->sin6_addr = sin6_dst->sin6_addr;
-		if (sc->gif_ro6.ro_rt) {
-			RTFREE(sc->gif_ro6.ro_rt);
-			sc->gif_ro6.ro_rt = NULL;
-		}
-#if 0
-		GIF2IFP(sc)->if_mtu = GIF_MTU;
-#endif
-	}
-
-	if (sc->gif_ro6.ro_rt == NULL) {
-		in6_rtalloc(&sc->gif_ro6, sc->gif_fibnum);
-		if (sc->gif_ro6.ro_rt == NULL) {
-			m_freem(m);
-			return ENETUNREACH;
-		}
-
-		/* if it constitutes infinite encapsulation, punt. */
-		if (sc->gif_ro.ro_rt->rt_ifp == ifp) {
-			m_freem(m);
-			return ENETUNREACH;	/*XXX*/
-		}
-#if 0
-		ifp->if_mtu = sc->gif_ro6.ro_rt->rt_ifp->if_mtu
-			- sizeof(struct ip6_hdr);
-#endif
-	}
-
-	m->m_flags &= ~(M_BCAST|M_MCAST);
-#ifdef IPV6_MINMTU
+	ip6->ip6_flow  |= htonl((uint32_t)ecn << 20);
+	ip6->ip6_nxt	= proto;
+	ip6->ip6_hlim	= V_ip6_gif_hlim;
 	/*
 	 * force fragmentation to minimum MTU, to avoid path MTU discovery.
 	 * it is too painful to ask for resend of inner packet, to achieve
 	 * path MTU discovery for encapsulated packets.
 	 */
-	error = ip6_output(m, 0, &sc->gif_ro6, IPV6_MINMTU, 0, NULL, NULL);
-#else
-	error = ip6_output(m, 0, &sc->gif_ro6, 0, 0, NULL, NULL);
-#endif
-
-	if (!(GIF2IFP(sc)->if_flags & IFF_LINK0) &&
-	    sc->gif_ro6.ro_rt != NULL) {
-		RTFREE(sc->gif_ro6.ro_rt);
-		sc->gif_ro6.ro_rt = NULL;
-	}
-
-	return (error);
+	return (ip6_output(m, 0, NULL, IPV6_MINMTU, 0, NULL, NULL));
 }
 
 int
 in6_gif_input(struct mbuf **mp, int *offp, int proto)
 {
 	struct mbuf *m = *mp;
-	struct ifnet *gifp = NULL;
+	struct ifnet *gifp;
 	struct gif_softc *sc;
 	struct ip6_hdr *ip6;
-	int af = 0;
-	u_int32_t otos;
+	uint8_t ecn;
 
-	ip6 = mtod(m, struct ip6_hdr *);
-
-	sc = (struct gif_softc *)encap_getarg(m);
+	sc = encap_getarg(m);
 	if (sc == NULL) {
 		m_freem(m);
 		IP6STAT_INC(ip6s_nogif);
-		return IPPROTO_DONE;
+		return (IPPROTO_DONE);
 	}
-
 	gifp = GIF2IFP(sc);
-	if (gifp == NULL || (gifp->if_flags & IFF_UP) == 0) {
-		m_freem(m);
-		IP6STAT_INC(ip6s_nogif);
-		return IPPROTO_DONE;
-	}
-
-	otos = ip6->ip6_flow;
-	m_adj(m, *offp);
-
-	switch (proto) {
-#ifdef INET
-	case IPPROTO_IPV4:
-	    {
-		struct ip *ip;
-		u_int8_t otos8;
-		af = AF_INET;
-		otos8 = (ntohl(otos) >> 20) & 0xff;
-		if (m->m_len < sizeof(*ip)) {
-			m = m_pullup(m, sizeof(*ip));
-			if (!m)
-				return IPPROTO_DONE;
-		}
-		ip = mtod(m, struct ip *);
-		if (ip_ecn_egress((gifp->if_flags & IFF_LINK1) ?
-				  ECN_ALLOWED : ECN_NOCARE,
-				  &otos8, &ip->ip_tos) == 0) {
-			m_freem(m);
-			return IPPROTO_DONE;
-		}
-		break;
-	    }
-#endif /* INET */
-#ifdef INET6
-	case IPPROTO_IPV6:
-	    {
-		struct ip6_hdr *ip6;
-		af = AF_INET6;
-		if (m->m_len < sizeof(*ip6)) {
-			m = m_pullup(m, sizeof(*ip6));
-			if (!m)
-				return IPPROTO_DONE;
-		}
+	if ((gifp->if_flags & IFF_UP) != 0) {
 		ip6 = mtod(m, struct ip6_hdr *);
-		if (ip6_ecn_egress((gifp->if_flags & IFF_LINK1) ?
-				   ECN_ALLOWED : ECN_NOCARE,
-				   &otos, &ip6->ip6_flow) == 0) {
-			m_freem(m);
-			return IPPROTO_DONE;
-		}
-		break;
-	    }
-#endif
-	case IPPROTO_ETHERIP:
-		af = AF_LINK;
-		break;
-
-	default:
-		IP6STAT_INC(ip6s_nogif);
+		ecn = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
+		m_adj(m, *offp);
+		gif_input(m, gifp, proto, ecn);
+	} else {
 		m_freem(m);
-		return IPPROTO_DONE;
+		IP6STAT_INC(ip6s_nogif);
 	}
-
-	gif_input(m, af, gifp);
-	return IPPROTO_DONE;
+	return (IPPROTO_DONE);
 }
 
 /*
@@ -380,19 +181,16 @@ static int
 gif_validate6(const struct ip6_hdr *ip6, struct gif_softc *sc,
     struct ifnet *ifp)
 {
-	struct sockaddr_in6 *src, *dst;
-
-	src = (struct sockaddr_in6 *)sc->gif_psrc;
-	dst = (struct sockaddr_in6 *)sc->gif_pdst;
 
+	GIF_RLOCK_ASSERT(sc);
 	/*
 	 * Check for address match.  Note that the check is for an incoming
 	 * packet.  We should compare the *source* address in our configuration
 	 * and the *destination* address of the packet, and vice versa.
 	 */
-	if (!IN6_ARE_ADDR_EQUAL(&src->sin6_addr, &ip6->ip6_dst) ||
-	    !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_src))
-		return 0;
+	if (!IN6_ARE_ADDR_EQUAL(&sc->gif_ip6hdr->ip6_src, &ip6->ip6_dst) ||
+	    !IN6_ARE_ADDR_EQUAL(&sc->gif_ip6hdr->ip6_dst, &ip6->ip6_src))
+		return (0);
 
 	/* martian filters on outer source - done in ip6_input */
 
@@ -410,29 +208,22 @@ gif_validate6(const struct ip6_hdr *ip6, struct gif_softc *sc,
 		rt = in6_rtalloc1((struct sockaddr *)&sin6, 0, 0UL,
 		    sc->gif_fibnum);
 		if (!rt || rt->rt_ifp != ifp) {
-#if 0
-			char ip6buf[INET6_ADDRSTRLEN];
-			log(LOG_WARNING, "%s: packet from %s dropped "
-			    "due to ingress filter\n", if_name(GIF2IFP(sc)),
-			    ip6_sprintf(ip6buf, &sin6.sin6_addr));
-#endif
 			if (rt)
 				RTFREE_LOCKED(rt);
-			return 0;
+			return (0);
 		}
 		RTFREE_LOCKED(rt);
 	}
 
-	return 128 * 2;
+	return (128 * 2);
 }
 
 /*
  * we know that we are in IFF_UP, outer address available, and outer family
  * matched the physical addr family.  see gif_encapcheck().
- * sanity check for arg should have been done in the caller.
  */
 int
-gif_encapcheck6(const struct mbuf *m, int off, int proto, void *arg)
+in6_gif_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
 {
 	struct ip6_hdr ip6;
 	struct gif_softc *sc;
@@ -440,31 +231,21 @@ gif_encapcheck6(const struct mbuf *m, int off, int proto, void *arg)
 
 	/* sanity check done in caller */
 	sc = (struct gif_softc *)arg;
+	GIF_RLOCK_ASSERT(sc);
 
-	/* LINTED const cast */
 	m_copydata(m, 0, sizeof(ip6), (caddr_t)&ip6);
 	ifp = ((m->m_flags & M_PKTHDR) != 0) ? m->m_pkthdr.rcvif : NULL;
-
-	return gif_validate6(&ip6, sc, ifp);
+	return (gif_validate6(&ip6, sc, ifp));
 }
 
 int
 in6_gif_attach(struct gif_softc *sc)
 {
-	sc->encap_cookie6 = encap_attach_func(AF_INET6, -1, gif_encapcheck,
-	    (void *)&in6_gif_protosw, sc);
-	if (sc->encap_cookie6 == NULL)
-		return EEXIST;
-	return 0;
-}
-
-int
-in6_gif_detach(struct gif_softc *sc)
-{
-	int error;
 
-	error = encap_detach(sc->encap_cookie6);
-	if (error == 0)
-		sc->encap_cookie6 = NULL;
-	return error;
+	KASSERT(sc->gif_ecookie == NULL, ("gif_ecookie isn't NULL"));
+	sc->gif_ecookie = encap_attach_func(AF_INET6, -1, gif_encapcheck,
+	    (void *)&in6_gif_protosw, sc);
+	if (sc->gif_ecookie == NULL)
+		return (EEXIST);
+	return (0);
 }