From 9ee84cddef0af6abc4d9dfc1cd142a36175a4e6a Mon Sep 17 00:00:00 2001
From: rwatson <rwatson@FreeBSD.org>
Date: Mon, 21 Apr 2008 12:06:41 +0000
Subject: With IPv4 raw sockets, read lock rather than write lock the inpcb
 when receiving or transmitting.

With IPv6 raw sockets, read lock rather than write lock the inpcb when
receiving.  Unfortunately, IPv6 source address selection appears to
require a write lock on the inpcb for the time being.

MFC after:	3 months
---
 sys/netinet/raw_ip.c   | 30 +++++++++++++++---------------
 sys/netinet6/raw_ip6.c | 10 +++++-----
 2 files changed, 20 insertions(+), 20 deletions(-)

(limited to 'sys')

diff --git a/sys/netinet/raw_ip.c b/sys/netinet/raw_ip.c
index ead5219..23ab1fe 100644
--- a/sys/netinet/raw_ip.c
+++ b/sys/netinet/raw_ip.c
@@ -155,7 +155,7 @@ raw_append(struct inpcb *last, struct ip *ip, struct mbuf *n)
 {
 	int policyfail = 0;
 
-	INP_WLOCK_ASSERT(last);
+	INP_RLOCK_ASSERT(last);
 
 #ifdef IPSEC
 	/* check AH/ESP integrity. */
@@ -209,10 +209,10 @@ rip_input(struct mbuf *m, int off)
 	ripsrc.sin_addr = ip->ip_src;
 	last = NULL;
 	LIST_FOREACH(inp, &ripcb, inp_list) {
-		INP_WLOCK(inp);
+		INP_RLOCK(inp);
 		if (inp->inp_ip_p && inp->inp_ip_p != proto) {
 	docontinue:
-			INP_WUNLOCK(inp);
+			INP_RUNLOCK(inp);
 			continue;
 		}
 #ifdef INET6
@@ -236,14 +236,14 @@ rip_input(struct mbuf *m, int off)
 			if (n != NULL)
 				(void) raw_append(last, ip, n);
 			/* XXX count dropped packet */
-			INP_WUNLOCK(last);
+			INP_RUNLOCK(last);
 		}
 		last = inp;
 	}
 	if (last != NULL) {
 		if (raw_append(last, ip, m) != 0)
 			ipstat.ips_delivered--;
-		INP_WUNLOCK(last);
+		INP_RUNLOCK(last);
 	} else {
 		m_freem(m);
 		ipstat.ips_noproto++;
@@ -278,7 +278,7 @@ rip_output(struct mbuf *m, struct socket *so, u_long dst)
 		if (m == NULL)
 			return(ENOBUFS);
 
-		INP_WLOCK(inp);
+		INP_RLOCK(inp);
 		ip = mtod(m, struct ip *);
 		ip->ip_tos = inp->inp_ip_tos;
 		if (inp->inp_flags & INP_DONTFRAG)
@@ -299,12 +299,12 @@ rip_output(struct mbuf *m, struct socket *so, u_long dst)
 			m_freem(m);
 			return(EMSGSIZE);
 		}
-		INP_WLOCK(inp);
+		INP_RLOCK(inp);
 		ip = mtod(m, struct ip *);
 		if (jailed(inp->inp_socket->so_cred)) {
 			if (ip->ip_src.s_addr !=
 			    htonl(prison_getip(inp->inp_socket->so_cred))) {
-				INP_WUNLOCK(inp);
+				INP_RUNLOCK(inp);
 				m_freem(m);
 				return (EPERM);
 			}
@@ -315,7 +315,7 @@ rip_output(struct mbuf *m, struct socket *so, u_long dst)
 		     && inp->inp_options)
 		    || (ip->ip_len > m->m_pkthdr.len)
 		    || (ip->ip_len < (ip->ip_hl << 2))) {
-			INP_WUNLOCK(inp);
+			INP_RUNLOCK(inp);
 			m_freem(m);
 			return EINVAL;
 		}
@@ -335,7 +335,7 @@ rip_output(struct mbuf *m, struct socket *so, u_long dst)
 
 	error = ip_output(m, inp->inp_options, NULL, flags,
 	    inp->inp_moptions, inp);
-	INP_WUNLOCK(inp);
+	INP_RUNLOCK(inp);
 	return error;
 }
 
@@ -851,13 +851,13 @@ rip_pcblist(SYSCTL_HANDLER_ARGS)
 	INP_INFO_RLOCK(&ripcbinfo);
 	for (inp = LIST_FIRST(ripcbinfo.ipi_listhead), i = 0; inp && i < n;
 	     inp = LIST_NEXT(inp, inp_list)) {
-		INP_WLOCK(inp);
+		INP_RLOCK(inp);
 		if (inp->inp_gencnt <= gencnt &&
 		    cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0) {
 			/* XXX held references? */
 			inp_list[i++] = inp;
 		}
-		INP_WUNLOCK(inp);
+		INP_RUNLOCK(inp);
 	}
 	INP_INFO_RUNLOCK(&ripcbinfo);
 	n = i;
@@ -865,7 +865,7 @@ rip_pcblist(SYSCTL_HANDLER_ARGS)
 	error = 0;
 	for (i = 0; i < n; i++) {
 		inp = inp_list[i];
-		INP_WLOCK(inp);
+		INP_RLOCK(inp);
 		if (inp->inp_gencnt <= gencnt) {
 			struct xinpcb xi;
 			bzero(&xi, sizeof(xi));
@@ -874,10 +874,10 @@ rip_pcblist(SYSCTL_HANDLER_ARGS)
 			bcopy(inp, &xi.xi_inp, sizeof *inp);
 			if (inp->inp_socket)
 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
-			INP_WUNLOCK(inp);
+			INP_RUNLOCK(inp);
 			error = SYSCTL_OUT(req, &xi, sizeof xi);
 		} else
-			INP_WUNLOCK(inp);
+			INP_RUNLOCK(inp);
 	}
 	if (!error) {
 		/*
diff --git a/sys/netinet6/raw_ip6.c b/sys/netinet6/raw_ip6.c
index c12bde1..b791872 100644
--- a/sys/netinet6/raw_ip6.c
+++ b/sys/netinet6/raw_ip6.c
@@ -154,10 +154,10 @@ rip6_input(struct mbuf **mp, int *offp, int proto)
 
 	INP_INFO_RLOCK(&ripcbinfo);
 	LIST_FOREACH(in6p, &ripcb, inp_list) {
-		INP_WLOCK(in6p);
+		INP_RLOCK(in6p);
 		if ((in6p->in6p_vflag & INP_IPV6) == 0) {
 docontinue:
-			INP_WUNLOCK(in6p);
+			INP_RUNLOCK(in6p);
 			continue;
 		}
 		if (in6p->in6p_ip6_nxt &&
@@ -207,7 +207,7 @@ docontinue:
 					sorwakeup(last->in6p_socket);
 				opts = NULL;
 			}
-			INP_WUNLOCK(last);
+			INP_RUNLOCK(last);
 		}
 		last = in6p;
 	}
@@ -220,7 +220,7 @@ docontinue:
 		ipsec6stat.in_polvio++;
 		ip6stat.ip6s_delivered--;
 		/* do not inject data into pcb */
-		INP_WUNLOCK(last);
+		INP_RUNLOCK(last);
 	} else
 #endif /* IPSEC */
 	if (last) {
@@ -237,7 +237,7 @@ docontinue:
 			rip6stat.rip6s_fullsock++;
 		} else
 			sorwakeup(last->in6p_socket);
-		INP_WUNLOCK(last);
+		INP_RUNLOCK(last);
 	} else {
 		rip6stat.rip6s_nosock++;
 		if (m->m_flags & M_MCAST)
-- 
cgit v1.1