| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Fri, 11 Mar 2022 09:25:09 -0800
From: greearb@...delatech.com
To: netdev@...r.kernel.org
Cc: dsahern@...il.com, Ben Greear <greearb@...delatech.com>
Subject: [PATCH] vrf/mcast: Fix mcast routing when using vrf.
From: Ben Greear <greearb@...delatech.com>
The problem case is where you have a VRF with 2 ports. Assume
eth1 points towards upstream and has default gateway, eth2 points
towards a local subnet.
User binds a UDP multicast socket to eth2 and attempts to send
multicast traffic on a non-local-multicast address (something other
than 224.0.0.x/24). There is no multicast router daemon in this case.
The flow through the kernel will start out with the correct oif (eth2),
but in this logic, the oif becomes the index of the vrf device instead:
/* update flow if oif or iif point to device enslaved to l3mdev */
l3mdev_update_flow(net, flowi4_to_flowi(flp));
After that, the mcast routing logic will choose the eth1 interface
as the output device, and the code in __mkroute_output will in the
end cause the frame to be sent out eth1 instead of the desired eth2.
To fix this, add special case logic to detect this in __mkroute_output
and instead use the user-specified port as the output device when
there is no specific mcast route available.
Signed-off-by: Ben Greear <greearb@...delatech.com>
---
net/ipv4/route.c | 25 +++++++++++++++++++++----
1 file changed, 21 insertions(+), 4 deletions(-)
diff --git a/net/ipv4/route.c b/net/ipv4/route.c
index f33ad1f383b6..722df8fcf417 100644
--- a/net/ipv4/route.c
+++ b/net/ipv4/route.c
@@ -2475,6 +2475,7 @@ int ip_route_input_rcu(struct sk_buff *skb, __be32 daddr, __be32 saddr,
/* called with rcu_read_lock() */
static struct rtable *__mkroute_output(const struct fib_result *res,
const struct flowi4 *fl4, int orig_oif,
+ struct net_device *orig_dev_out,
struct net_device *dev_out,
unsigned int flags)
{
@@ -2516,12 +2517,26 @@ static struct rtable *__mkroute_output(const struct fib_result *res,
flags &= ~RTCF_LOCAL;
else
do_cache = false;
- /* If multicast route do not exist use
- * default one, but do not gateway in this case.
+ /* If multicast route does not exist use
+ * default one, but do not use gateway in this case.
* Yes, it is hack.
*/
- if (fi && res->prefixlen < 4)
+ if (fi && res->prefixlen < 4) {
+ struct net *net = dev_net(dev_out);
+
fi = NULL;
+
+ if (orig_oif && orig_dev_out &&
+ dev_out->ifindex != orig_oif &&
+ netif_index_is_l3_master(net, fl4->flowi4_oif)) {
+ /* vrf overwrites the original flowi4_oif for
+ * member network devices. In that case,
+ * lets use the user-specified oif instead of
+ * a default route.
+ */
+ dev_out = orig_dev_out;
+ }
+ }
} else if ((type == RTN_LOCAL) && (orig_oif != 0) &&
(orig_oif != dev_out->ifindex)) {
/* For local routes that require a particular output interface
@@ -2630,6 +2645,7 @@ struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *fl4,
const struct sk_buff *skb)
{
struct net_device *dev_out = NULL;
+ struct net_device *orig_dev_out = NULL;
int orig_oif = fl4->flowi4_oif;
unsigned int flags = 0;
struct rtable *rth;
@@ -2785,12 +2801,13 @@ struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *fl4,
goto make_route;
}
+ orig_dev_out = dev_out;
fib_select_path(net, res, fl4, skb);
dev_out = FIB_RES_DEV(*res);
make_route:
- rth = __mkroute_output(res, fl4, orig_oif, dev_out, flags);
+ rth = __mkroute_output(res, fl4, orig_oif, orig_dev_out, dev_out, flags);
out:
return rth;
--
2.20.1
Powered by blists - more mailing lists