On Fri, Jul 21, 2017 at 2:53 PM, Roopa Prabhu [off-list ref] wrote:

On Fri, Jul 21, 2017 at 11:42 AM, Cong Wang [off-list ref] wrote:

quoted

On Thu, Jul 20, 2017 at 8:23 AM, Hangbin Liu [off-list ref] wrote:

quoted

2017-07-20 23:06 GMT+08:00 Hangbin Liu [off-list ref]:

quoted

quoted

+++ b/net/ipv6/route.c

@@ -3637,12 +3637,6 @@ static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
                dst = ip6_route_lookup(net, &fl6, 0);

        rt = container_of(dst, struct rt6_info, dst);
-       if (rt->dst.error) {
-               err = rt->dst.error;
-               ip6_rt_put(rt);
-               goto errout;
-       }

hmm... or instead of remove this check, should we check all the entry? Like
if ((rt->dst.error && rt != net->ipv6.ip6_null_entry && rt !=

                                                    ^^ mistake here

quoted

net->ipv6.ip6_blk_hole_entry) ||
     rt == net->ipv6.ip6_null_entry )

Sorry,  there should be no need to check ip6_null_entry since the
error is already
-ENETUNREACH. So how about

Hmm? All of these 3 entries have error set, right??
So we should only check dst.error...

removing the check seems ok to me. We can also make the check
conditional to fibmatch code only
to eliminate any change in behavior introduced by fibmatch.

ie if (fibmatch && rt->dst.error).

Hangbin, can you also pls check that fibmatch works ok for such routes
with your patch applied ?.

ip netns exec client ip -6 route get fibmatch 2003::1     (with latest iproute2)

thank you.

I tried this with your patch and this works for the fibmatch case as well.