Re: [PATCH net-next] ipv6: host should probe those non-reachable router

[Hannes Frederic Sowa <hannes@stressinduktion.org>]

Hi Duan!

On Fri, Nov 01, 2013 at 10:47:15AM +0800, Duan Jiong wrote:
> 
> rather than those routers that the return value of rt6_score_route()
> is RT6_NUD_SUCCEED.
> 
> Signed-off-by: Duan Jiong <duanj.fnst@cn.fujitsu.com>
> ---
>  net/ipv6/route.c | 5 ++---
>  1 file changed, 2 insertions(+), 3 deletions(-)
> 
> diff --git a/net/ipv6/route.c b/net/ipv6/route.c
> index 1ac0b6e..b7bf5c7 100644
> --- a/net/ipv6/route.c
> +++ b/net/ipv6/route.c
> @@ -623,12 +623,11 @@ static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
>  		match_do_rr = true;
>  		m = 0; /* lowest valid score */
>  	} else if (m < 0) {
> +		if (strict & RT6_LOOKUP_F_REACHABLE)
> +			rt6_probe(rt);
>  		goto out;
>  	}
>  
> -	if (strict & RT6_LOOKUP_F_REACHABLE)
> -		rt6_probe(rt);
> -

I don't think this is correct. m < 0 are the routers we really don't want to
use, e.g. because we get a failure signaled from rt6_check_dev. E.g. we want
to query a route where a specified interface id is given.

In case router preference is enabled we mostly give back RT6_NUD_SUCCEED for
routes which are somewhat reachable. Really broken routers in NUD_FAILED state
are waiting for cleanup and recreation. We also do not accept new NAs on
neighbours in NUD_FAILED state IIRC.

Greetings,

  Hannes