Re: IPv6 prefix route lifetime

[Ryota Ozaki <ozaki-r%netbsd.org@localhost>]

On Tue, Aug 9, 2016 at 5:56 PM, Roy Marples <roy%marples.name@localhost> wrote:
> On 08/08/2016 07:49, Ryota Ozaki wrote:
>> I see the following errors.
>
> test rump tests currently panic on yesterdays buildable current which
> makes fixing it hard.

Yes. So I tested your patch at top of your commit of RTF_CONNECTED:
  https://github.com/IIJ-NetBSD/netbsd-src/commit/81196b0d8df1e0c452c058850256b1e883cfb108

>
> The only difference is that prefix routes are now generated using rtinit
> rather than a custom rtrequest call which places the LL-addr as the
> gateway instead of the address. This aligns it with IPv4 subnet routes
> (and is a lot less code to worry about) so I would rather try and fix
> the tests.
>
>>
>> === route_flags_lo6 ===
>> tc-so:Executing command [ /bin/sh -c rump.netstat -rn -f inet6 | grep ^'::1' ]
>> tc-se:Fail: regexp  UH  not in stdout
>> tc-se:::1                                     lo0
>>       UHl         -        -  33624  lo0
>
> This test is easily fixable - while the flags are now different, it
> doesn't really matter.

Oh sure. UHl is proper in favor of introducing RTF_LOCAL.
I think the change for it should be a separated commit if possible.

  ozaki-r

>
>> === recursiveipv6overipv4 and recursiveipv6overipv6 ====
>> tc-so:Executing command [ /bin/sh -c env
>> LD_PRELOAD=/usr/lib/librumphijack.so RUMPHIJACK=sysctl=yes dmesg ]
>> tc-se:Fail: regexp gif0: recursively called too many times not in stdout
>
> I don't understand this error.
> Does it imply there is no output to parse?
> I will try looking into why this happens in the gif driver.
>
> Roy