[PATCH net-next] ipv6: check mc_forwarding before calling ip6mr_get_route()

[PATCH net-next] ipv6: check mc_forwarding before calling ip6mr_get_route()

[Wenbo Sun]

When CONFIG_IPV6_MROUTE is enabled, rt6_fill_node() unconditionally calls
ip6mr_get_route() whenever an input interface is specified and the
destination address is multicast.

However, if multicast routing is not active (i.e. no multicast routing
daemon is running and mc_forwarding is 0), there is no matching (S,G)
multicast routing cache entry, and skb->dev is NULL in the synthetic route
query skb allocated by inet6_rtm_getroute(). As a result, ip6mr_get_route()
returns -ENODEV, which causes rt6_fill_node() to jump to nla_put_failure
and return -EMSGSIZE to userspace ("RTNETLINK answers: Message too long").

In contrast, IPv4 counterpart rt_fill_info() checks
IPV4_DEVCONF_ALL_RO(net, MC_FORWARDING) before attempting to query
ipmr_get_route(), and falls back to emitting RTA_IIF and normal local
delivery route attributes when multicast routing is not enabled.

This discrepancy was identified during code audit of multicast route get
handling. The failure was confirmed on a UML kernel: querying an IPv6
multicast route with an input interface ("ip -6 route get <mcast_addr>
from <src_addr> ...") failed with -EMSGSIZE on an unpatched kernel.

Align IPv6 behavior with IPv4 by checking whether mc_forwarding is active
(atomic_read(&net->ipv6.devconf_all->mc_forwarding)) before calling
ip6mr_get_route(). When mc_forwarding is 0, rt6_fill_node() proceeds to
output RTA_IIF and normal route attributes.

Tested:
- On a UML kernel with CONFIG_IPV6_MROUTE=y, verified that
  "ip -6 route get ff0e::1 from 2001:db8::3 ..." with an input interface
  now succeeds and outputs the route instead of failing with -EMSGSIZE.
- Verified that IPv6 unicast route get and multicast route get without
  an input interface continue to work as expected.
- Built with W=1 on UML, x86_64 allmodconfig, and x86_64 allyesconfig with
  zero new warnings or errors.

Signed-off-by: Wenbo Sun <sunwenbo@cmss.chinamobile.com>
---
 net/ipv6/route.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/net/ipv6/route.c b/net/ipv6/route.c
index 2eb884558c24..6dd4a70f6490 100644
--- a/net/ipv6/route.c
+++ b/net/ipv6/route.c
@@ -5862,7 +5862,8 @@ static int rt6_fill_node(struct net *net, struct sk_buff *skb,
 #endif
 	if (iif) {
 #ifdef CONFIG_IPV6_MROUTE
-		if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
+		if (ipv6_addr_is_multicast(&rt6_dst->addr) &&
+		    atomic_read(&net->ipv6.devconf_all->mc_forwarding)) {
 			int err = ip6mr_get_route(net, skb, rtm, portid);
 
 			if (err == 0)
-- 
2.33.0

Re: [PATCH net-next] ipv6: check mc_forwarding before calling ip6mr_get_route()

[Hangbin Liu]

Hi Wenbo,
On Wed, Nov 11, 2026 at 05:45:28PM +0800, Wenbo Sun wrote:
> When CONFIG_IPV6_MROUTE is enabled, rt6_fill_node() unconditionally calls
> ip6mr_get_route() whenever an input interface is specified and the
> destination address is multicast.
> 
> However, if multicast routing is not active (i.e. no multicast routing
> daemon is running and mc_forwarding is 0), there is no matching (S,G)
> multicast routing cache entry, and skb->dev is NULL in the synthetic route
> query skb allocated by inet6_rtm_getroute(). As a result, ip6mr_get_route()
> returns -ENODEV, which causes rt6_fill_node() to jump to nla_put_failure
> and return -EMSGSIZE to userspace ("RTNETLINK answers: Message too long").
> 
> In contrast, IPv4 counterpart rt_fill_info() checks
> IPV4_DEVCONF_ALL_RO(net, MC_FORWARDING) before attempting to query
> ipmr_get_route(), and falls back to emitting RTA_IIF and normal local
> delivery route attributes when multicast routing is not enabled.
> 
> This discrepancy was identified during code audit of multicast route get
> handling. The failure was confirmed on a UML kernel: querying an IPv6
> multicast route with an input interface ("ip -6 route get <mcast_addr>
> from <src_addr> ...") failed with -EMSGSIZE on an unpatched kernel.
> 
> Align IPv6 behavior with IPv4 by checking whether mc_forwarding is active
> (atomic_read(&net->ipv6.devconf_all->mc_forwarding)) before calling
> ip6mr_get_route(). When mc_forwarding is 0, rt6_fill_node() proceeds to
> output RTA_IIF and normal route attributes.
> 
> Tested:
> - On a UML kernel with CONFIG_IPV6_MROUTE=y, verified that
>   "ip -6 route get ff0e::1 from 2001:db8::3 ..." with an input interface
>   now succeeds and outputs the route instead of failing with -EMSGSIZE.
> - Verified that IPv6 unicast route get and multicast route get without
>   an input interface continue to work as expected.
> - Built with W=1 on UML, x86_64 allmodconfig, and x86_64 allyesconfig with
>   zero new warnings or errors.
> 
> Signed-off-by: Wenbo Sun <sunwenbo@cmss.chinamobile.com>
> ---
>  net/ipv6/route.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/net/ipv6/route.c b/net/ipv6/route.c
> index 2eb884558c24..6dd4a70f6490 100644
> --- a/net/ipv6/route.c
> +++ b/net/ipv6/route.c
> @@ -5862,7 +5862,8 @@ static int rt6_fill_node(struct net *net, struct sk_buff *skb,
>  #endif
>  	if (iif) {
>  #ifdef CONFIG_IPV6_MROUTE
> -		if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
> +		if (ipv6_addr_is_multicast(&rt6_dst->addr) &&
> +		    atomic_read(&net->ipv6.devconf_all->mc_forwarding)) {
>  			int err = ip6mr_get_route(net, skb, rtm, portid);
>  
>  			if (err == 0)

In commit fd61c6ba313d ("net: ipv6: remove nowait arg to rt6_fill_node"),
David Ahern removed the nowait arg. Should we still return immediately
after err == 0? Also later if (err < 0), it still goto nla_put_failure
and return -EMSGSIZE. Maybe we can carry the err to the end and return
a more meaningful reason?

Anyway the added checking makes sense to me.

Reviewed-by: Hangbin Liu <liuhangbin@kylinos.cn>