On 07/08/16 23:35, Alexander Bluhm wrote:
> On Tue, Jun 28, 2016 at 08:30:16AM +0200, Martin Pieuchot wrote:
>> With this diff if your next hop becomes invalid after being cached you'll
>> also need two ICMP6_PACKET_TOO_BIG to restore the MTU, is this wanted?
>
> No, a single ICMP6_PACKET_TOO_BIG
On Tue, Jun 28, 2016 at 08:30:16AM +0200, Martin Pieuchot wrote:
> With this diff if your next hop becomes invalid after being cached you'll
> also need two ICMP6_PACKET_TOO_BIG to restore the MTU, is this wanted?
No, a single ICMP6_PACKET_TOO_BIG packet is enough. I have written
a test that
On 25/06/16(Sat) 01:03, Alexander Bluhm wrote:
> Hi,
>
> Path MTU discovery in IPv6 is slightly broken. I takes two
> ICMP6_PACKET_TOO_BIG packets to create and change the dynamic route.
>
> When the first ICMP6 packet arrives, a dynamic route gets created
> in icmp6_mtudisc_clone() and
Hi,
Path MTU discovery in IPv6 is slightly broken. I takes two
ICMP6_PACKET_TOO_BIG packets to create and change the dynamic route.
When the first ICMP6 packet arrives, a dynamic route gets created
in icmp6_mtudisc_clone() and icmp6_mtudisc_update() sets its MTU
value. During notify