[PATCH] ipv6: Fix default route failover when CONFIG_IPV6_ROUTER_PREF=n

I believe this commit from 2008 was incorrect:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commitdiff;h=398bcbebb6f721ac308df1e3d658c0029bb74503

When CONFIG_IPV6_ROUTER_PREF is disabled, the kernel should follow
RFC4861 section 6.3.6: if no route is NUD_VALID, then traffic should be
sprayed across all routers (indirectly triggering NUD) until one of them
becomes NUD_VALID.

However, the following experiment demonstrates that this does not work:

1) Connect to an IPv6 network.
2) Change the router's MAC (and link-local) address.

The kernel will lock onto the first router and never try the new one, even
if the first becomes unreachable.  This patch fixes the problem by
allowing rt6_check_neigh() to return 0; if all routers return 0, then
rt6_select() will fall back to round-robin behavior.

This patch should have no effect when CONFIG_IPV6_ROUTER_PREF=y.

Note that rt6_check_neigh() is only used in a boolean context, so the
presence of both "m = 1" and "m = 2" was irrelevant and confusing.

Signed-off-by: Paul Marks <redacted>
---
 net/ipv6/route.c |   13 +++++--------
 1 files changed, 5 insertions(+), 8 deletions(-)

diff --git a/net/ipv6/route.c b/net/ipv6/route.c
index b1e6cf0..60b2995 100644
--- a/net/ipv6/route.c
+++ b/net/ipv6/route.c

@@ -490,7 +490,7 @@ static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 static inline int rt6_check_neigh(struct rt6_info *rt)
 {
 	struct neighbour *neigh;
-	int m;
+	int m = 0;
 
 	neigh = rt->n;
 	if (rt->rt6i_flags & RTF_NONEXTHOP ||

@@ -499,16 +499,13 @@ static inline int rt6_check_neigh(struct rt6_info *rt)
 	else if (neigh) {
 		read_lock_bh(&neigh->lock);
 		if (neigh->nud_state & NUD_VALID)
-			m = 2;
+			m = 1;
 #ifdef CONFIG_IPV6_ROUTER_PREF
-		else if (neigh->nud_state & NUD_FAILED)
-			m = 0;
-#endif
-		else
+		else if (!(neigh->nud_state & NUD_FAILED))
 			m = 1;
+#endif
 		read_unlock_bh(&neigh->lock);
-	} else
-		m = 0;
+	}
 	return m;
 }
 

-- 
1.7.7.3

From: Paul Marks <redacted>
Date: Sun,  2 Dec 2012 02:00:21 -0800

I believe this commit from 2008 was incorrect:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commitdiff;h=398bcbebb6f721ac308df1e3d658c0029bb74503

When CONFIG_IPV6_ROUTER_PREF is disabled, the kernel should follow
RFC4861 section 6.3.6: if no route is NUD_VALID, then traffic should be
sprayed across all routers (indirectly triggering NUD) until one of them
becomes NUD_VALID.

However, the following experiment demonstrates that this does not work:

1) Connect to an IPv6 network.
2) Change the router's MAC (and link-local) address.

The kernel will lock onto the first router and never try the new one, even
if the first becomes unreachable.  This patch fixes the problem by
allowing rt6_check_neigh() to return 0; if all routers return 0, then
rt6_select() will fall back to round-robin behavior.

This patch should have no effect when CONFIG_IPV6_ROUTER_PREF=y.

Note that rt6_check_neigh() is only used in a boolean context, so the
presence of both "m = 1" and "m = 2" was irrelevant and confusing.

Signed-off-by: Paul Marks <redacted>

The values '1' and '2' used to influence rt6_score_route() in different
ways.

Please change rt6_check_neigh() to return a "bool" if you are going to
do this.

Thanks.

I believe this commit from 2008 was incorrect:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commitdiff;h=398bcbebb6f721ac308df1e3d658c0029bb74503

When CONFIG_IPV6_ROUTER_PREF is disabled, the kernel should follow
RFC4861 section 6.3.6: if no route is NUD_VALID, then traffic should be
sprayed across all routers (indirectly triggering NUD) until one of them
becomes NUD_VALID.

However, the following experiment demonstrates that this does not work:

1) Connect to an IPv6 network.
2) Change the router's MAC (and link-local) address.

The kernel will lock onto the first router and never try the new one, even
if the first becomes unreachable.  This patch fixes the problem by
allowing rt6_check_neigh() to return 0; if all routers return 0, then
rt6_select() will fall back to round-robin behavior.

This patch should have no effect when CONFIG_IPV6_ROUTER_PREF=y.

Note that rt6_check_neigh() is only used in a boolean context, so I've
changed its return type accordingly.

Signed-off-by: Paul Marks <redacted>
---
Changelog since v1:
- Changed return type from 'int' to 'bool'.

 net/ipv6/route.c |   24 ++++++++++--------------
 1 files changed, 10 insertions(+), 14 deletions(-)

diff --git a/net/ipv6/route.c b/net/ipv6/route.c
index b1e6cf0..35f2815 100644
--- a/net/ipv6/route.c
+++ b/net/ipv6/route.c

@@ -487,35 +487,32 @@ static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 	return 0;
 }
 
-static inline int rt6_check_neigh(struct rt6_info *rt)
+static inline bool rt6_check_neigh(struct rt6_info *rt)
 {
 	struct neighbour *neigh;
-	int m;
+	bool ret = false;
 
 	neigh = rt->n;
 	if (rt->rt6i_flags & RTF_NONEXTHOP ||
 	    !(rt->rt6i_flags & RTF_GATEWAY))
-		m = 1;
+		ret = true;
 	else if (neigh) {
 		read_lock_bh(&neigh->lock);
 		if (neigh->nud_state & NUD_VALID)
-			m = 2;
+			ret = true;
 #ifdef CONFIG_IPV6_ROUTER_PREF
-		else if (neigh->nud_state & NUD_FAILED)
-			m = 0;
+		else if (!(neigh->nud_state & NUD_FAILED))
+			ret = true;
 #endif
-		else
-			m = 1;
 		read_unlock_bh(&neigh->lock);
-	} else
-		m = 0;
-	return m;
+	}
+	return ret;
 }
 
 static int rt6_score_route(struct rt6_info *rt, int oif,
 			   int strict)
 {
-	int m, n;
+	int m;
 
 	m = rt6_check_dev(rt, oif);
 	if (!m && (strict & RT6_LOOKUP_F_IFACE))

@@ -523,8 +520,7 @@ static int rt6_score_route(struct rt6_info *rt, int oif,
 #ifdef CONFIG_IPV6_ROUTER_PREF
 	m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 #endif
-	n = rt6_check_neigh(rt);
-	if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
+	if (!rt6_check_neigh(rt) && (strict & RT6_LOOKUP_F_REACHABLE))
 		return -1;
 	return m;
 }

-- 
1.7.7.3

From: Paul Marks <redacted>
Date: Mon,  3 Dec 2012 12:26:54 -0800

I believe this commit from 2008 was incorrect:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commitdiff;h=398bcbebb6f721ac308df1e3d658c0029bb74503

When CONFIG_IPV6_ROUTER_PREF is disabled, the kernel should follow
RFC4861 section 6.3.6: if no route is NUD_VALID, then traffic should be
sprayed across all routers (indirectly triggering NUD) until one of them
becomes NUD_VALID.

However, the following experiment demonstrates that this does not work:

1) Connect to an IPv6 network.
2) Change the router's MAC (and link-local) address.

The kernel will lock onto the first router and never try the new one, even
if the first becomes unreachable.  This patch fixes the problem by
allowing rt6_check_neigh() to return 0; if all routers return 0, then
rt6_select() will fall back to round-robin behavior.

This patch should have no effect when CONFIG_IPV6_ROUTER_PREF=y.

Note that rt6_check_neigh() is only used in a boolean context, so I've
changed its return type accordingly.

Signed-off-by: Paul Marks <redacted>

Applied to net-next, thanks Paul.