On 9/6/18 6:57 AM, Hangbin Liu wrote:
> When receive a redirect message and call rt6_do_redirect(), we allocate
> a new rt6_info and set new flags and gateway info, but not update these
> info to fib6_info.
>
> Then if a user try to get the route info via `ip route get`, he will still
> get the old default gateway, because inet6_rtm_getroute() get gateway info
> from fib6_info.
>
> Fixes: 23fb93a4d3f11 ("net/ipv6: Cleanup exception and cache route handling")
> Reported-by: Jianlin Shi <ji...@redhat.com>
> Signed-off-by: Hangbin Liu <liuhang...@gmail.com>
> ---
> net/ipv6/route.c | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/net/ipv6/route.c b/net/ipv6/route.c
> index 18e00ce..3d367c9 100644
> --- a/net/ipv6/route.c
> +++ b/net/ipv6/route.c
> @@ -3446,6 +3446,10 @@ static void rt6_do_redirect(struct dst_entry *dst,
> struct sock *sk, struct sk_bu
> goto out;
> }
>
> + /* Update fib6_info from rt6_info */
> + from->fib6_flags = rt->rt6i_flags;
> + from->fib6_nh.nh_gw = rt->rt6i_gateway;
> +
> netevent.old = &rt->dst;
> netevent.new = &nrt->dst;
> netevent.daddr = &msg->dest;
>
Only an exception should be inserted - and it is. The original route
should not be updated.
The code prior to the fib6_info did not update the actual FIB entry, and
the IPv4 code does not update the original route.
