Re: [PATCH net-next] ipv6: export IP6_RT_PRIO_* to userland

[David Miller <davem@davemloft.net>]

From: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Date: Fri, 09 Nov 2012 10:43:34 +0100

> Le 05/11/2012 20:24, Nicolas Dichtel a écrit :
>> Le 05/11/2012 18:43, David Miller a écrit :
>>> From: Nicolas Dichtel <nicolas.dichtel@6wind.com>
>>> Date: Mon, 05 Nov 2012 18:22:39 +0100
>>>
>>>> Le 05/11/2012 18:00, David Miller a écrit :
>>>>> From: Nicolas Dichtel <nicolas.dichtel@6wind.com>
>>>>> Date: Mon,  5 Nov 2012 16:28:18 +0100
>>>>>
>>>>>> The kernel uses some default metric when routes are managed. For
>>>>>> example, a
>>>>>> static route added with a metric set to 0 is inserted in the kernel
>>>>>> with
>>>>>> metric 1024 (IP6_RT_PRIO_USER).
>>>>>> It is useful for routing daemons to know these values, to be able to
>>>>>> set routes
>>>>>> without interfering with what the kernel does.
>>>>>>
>>>>>> Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
>>>>>
>>>>> But these belong in the libc headers anyways.
>>>>>
>>>>> If we haven't provided them for so long, there's no real value of
>>>>> adding them now.
>>>>>
>>>> But how can a daemon know which default values are used? If it wants
>>>> to add a route with metric = default + x, it should first add it with
>>>> metric = 0, check the result and then change the metric.
>>>
>>> What do we do on the ipv4 side and how do daemons cope in that
>>> situation?
>>>
>> In IPv4, there is no such default metric. If you add a route with
>> metric X, it
>> remains X in the kernel, even if it's 0.
> So we need to duplicate these values in the daemon?
> 

I'm without real internet access again for the past 2 days and simply
too backlogged to even continue considering this issue.

Please be patient.