On Tue, 2008-09-30 at 10:35 -0400, Neil Horman wrote:

On Tue, Sep 30, 2008 at 06:17:12PM +0400, Denis V. Lunev wrote:

quoted

The patch is wrong in respect to
  for_each_net
usage.

It should be used under rtnl. It is not held at the moment :(

Thank you, I'll be sure to take that into consideration when we determine what
the proper algorithm to implement here is.

Although, I used the rt_secret_rebuild timer function as my guide here, and it
doesn't hold the rtnl lock either, or is that because its running in a softirq
context?

no. There is no lock as there is no iteration over net namespace list.
Each namespace has its own timer.

The simplest approach I see right now is to re-iterate over current dst
cache chain and mark only those namespaces. This is at least fair as
only one namespace can be attacked (not entire node).

All locking for this is in place.

Regards,
	Den

Neil

quoted

Regards,
	Den

On Mon, 2008-09-29 at 15:12 -0400, Neil Horman wrote:

quoted

Hey all-
	We currently have the ability to disable our route cache secret interval
rebuild timer (by setting it to zero), but if we do that its possible for an
attacker (if they guess our route cache hash secret, to fill our system with
routes that all hash to the same bucket, destroying our performance.  This patch
provides a backstop for that issues.  In the event that our rebuild interval is
disabled (or very large), if any hash chain exceeds ip_rt_gc_elasticity, we do
an emergency hash rebuild.  During the hash rebuild we:
1) warn the user of the emergency
2) disable the rebuild timer
3) invalidate the route caches
4) re-enable the rebuild timer with its old value

Regards
Neil

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>


 route.c |   36 +++++++++++++++++++++++++++++++++++-
 1 file changed, 35 insertions(+), 1 deletion(-)

diff --git a/net/ipv4/route.c b/net/ipv4/route.c
index 6ee5354..b95e02a 100644
--- a/net/ipv4/route.c
+++ b/net/ipv4/route.c

@@ -145,7 +145,7 @@ static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst);
 static void		 ipv4_link_failure(struct sk_buff *skb);
 static void		 ip_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
 static int rt_garbage_collect(struct dst_ops *ops);
-
+static void rt_emergency_hash_rebuild(void);
 
 static struct dst_ops ipv4_dst_ops = {
 	.family =		AF_INET,

@@ -1056,6 +1056,15 @@ restart:
 			*candp = cand->u.dst.rt_next;
 			rt_free(cand);
 		}
+	} else if (chain_length > ip_rt_gc_elasticity) {
+		/*
+		 * We didn't find a route entry we could safely free,
+		 * yet our chain length is over our elasticity value.
+		 * Someone may have guessed our hash secret and is artifically
+		 * filling up our route cache.  Lets do an emergency rebuild
+		 * to be safe
+		 */
+		rt_emergency_hash_rebuild();
 	}
 
 	/* Try to bind route to arp only if it is output

@@ -2946,6 +2955,31 @@ static void rt_secret_reschedule(int old)
 	rtnl_unlock();
 }
 
+static void rt_secret_rebuild_oneshot(void) {
+	struct net *net;
+	int old_interval = ip_rt_secret_interval;
+
+	ip_rt_secret_interval = 0;
+
+	rt_secret_reschedule(old_interval);
+
+	for_each_net(net) 
+		rt_cache_invalidate(net);
+
+	ip_rt_secret_interval = old_interval;
+
+	rt_secret_reschedule(0);
+}
+
+static void rt_emergency_hash_rebuild(void) {
+	if (net_ratelimit()) {
+		printk(KERN_WARNING "Route hash chain too long!\n");
+		printk(KERN_WARNING "Adjust your secret_interval!\n");
+	}
+
+	rt_secret_rebuild_oneshot();
+}
+
 static int ipv4_sysctl_rt_secret_interval(ctl_table *ctl, int write,
 					  struct file *filp,
 					  void __user *buffer, size_t *lenp,