[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20170725000849.GD2938@leo.usersys.redhat.com>
Date: Tue, 25 Jul 2017 08:08:49 +0800
From: Hangbin Liu <liuhangbin@...il.com>
To: Cong Wang <xiyou.wangcong@...il.com>
Cc: network dev <netdev@...r.kernel.org>,
Roopa Prabhu <roopa@...ulusnetworks.com>
Subject: Re: [PATCH net] ipv6: no need to return rt->dst.error if it is not
null entry.
On Mon, Jul 24, 2017 at 12:57:43PM -0700, Cong Wang wrote:
> On Sun, Jul 23, 2017 at 8:09 PM, Hangbin Liu <liuhangbin@...il.com> wrote:
> > Do we still need this net->ipv6.ip6_null_entry check? How about remove all
> > the checks?
>
> I believe you only need to check for rt->dst.error, no need to check against
> NULL or ip6_null_entry.
>
> Take a look at other ip6_route_lookup() callers.
Yes, I saw it. That why I send v2 patch to check both rt->dst.error and
ip6_null_entry.
The question is the other two caller are rpfilter_lookup_reverse6() and
nft_fib6_eval(). From the code it looks these two caller only care about
device match.
if (rt->rt6i_idev->dev == dev || (flags & XT_RPFILTER_LOOSE))
ret = true;
And the device would be lo if it is ip6_null_entry. So they just discard it.
I'm not familiar with netfilter, Please correct me if I make any mistake.
But what we want in inet6_rtm_getroute() and rt6_dump_route() is to
get/dump the route info. So we should get the info even it's unreachable or
prohibit.
That's why I think we should remove both rt->dst.error and ip6_null_entry
check in inet6_rtm_getroute(). And even further, remove the ip6_null_entry
check in rt6_dump_route().
What do you think?
Thanks
Hangbin
Powered by blists - more mailing lists