On Wed, Jul 26, 2017 at 10:18 AM, David Ahern <dsah...@gmail.com> wrote:
> On 7/25/17 1:32 AM, Hangbin Liu wrote:
>> On Mon, Jul 24, 2017 at 09:28:07PM -0600, David Ahern wrote:
>>> On 7/24/17 6:08 PM, Hangbin Liu wrote:
>>>> 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().
>>>
>>> git blame net/ipv6/route.c
>>>
>>> find the commits and review.
>>
>> Hi David,
>>
>> Sorry, would you like to give me more hints?
>>
>> I saw your commit 1f17e2f2c8a8 ("net: ipv6: ignore null_entry on route
>> dumps"). But I think this issue has been fixed by
>>
>> 2f460933f58e ("ipv6: initialize route null entry in addrconf_init()") and
>> 242d3a49a2a1 ("ipv6: reorder ip6_route_dev_notifier after ipv6_dev_notf")
>>
>> I use a reproducer which add unreachable route in netns with lo down. And I
>> could not trigger Panic in the fixed kernel. That's why I think we could
>> remove ip6_null_entry check in rt6_dump_route().
>
> Understood. Cong's patch to fix the intialization order (lo device
> before route init) makes sure the idev is not null. That said, the
> null_entry route is internal ipv6 logic and is not a route entry to be
> returned to userspace.
agreed...so looks like the check in v3 should be
+ if ( rt == net->ipv6.ip6_null_entry ||
+ (rt->dst.error &&
+ #ifdef CONFIG_IPV6_MULTIPLE_TABLES
+ rt != net->ipv6.ip6_prohibit_entry &&
+ rt != net->ipv6.ip6_blk_hole_entry &&
+#endif
+ )) {
err = rt->dst.error;
ip6_rt_put(rt);
goto errout;
