Re: [PATCH net-next 5/5] ipv6: Compute multipath hash for forwarded ICMP errors from offending packet

From: Hannes Frederic Sowa
Date: Tue Nov 01 2016 - 13:01:24 EST

On 01.11.2016 17:39, David Miller wrote:
> From: Hannes Frederic Sowa <hannes@xxxxxxxxxxxxxxxxxxx>
> Date: Tue, 1 Nov 2016 17:27:56 +0100
>> On 01.11.2016 16:35, David Miller wrote:
>>> I have a really hard time accepting a "fix" that depends upon behavior
>>> that the Linux ipv6 stack doesn't even have.
>> We actually support this feature:
> But it is forbidden when the sysctl I mentioned is set, which is the
> default.
> I'm talking about default behavior, which is to not reflect.

Oh, yes, understood.

I think we can flip this sysctl by default to off: current default
kernel config actually generates flow labels on its own, so the
description of this sysctl is violated by default anyway, as it doesn't
preserve the uniqueness anymore.