every 2nd echo-request malformed when ping -s >4067

[Adrian Chadd]

On 24 October 2012 05:31, Harald Schmalzbauer <h.schmalzbauer at omnilan.de> wrote:
>  Hello,
>
> while checking new mtu9k-setup, I discovered that ping has some odd
> behaviour.
> If I use payloadsize > 4067, every 2nd icmp-echo-request seems to be
> malformed:

Is this on -HEAD?

> ping -s 4068 -D 10.5.49.65
>
> 1st: 12:21:09.048447 IP 10.5.49.126 > 10.5.49.65: ICMP echo request, id
> 46597, seq 0, length 4076
>         0x0000:  4500 1000 0f2d 4000 4001 a507 0a05 317e
>
> 2nd: 12:21:10.052891 IP 10.5.49.126 > 10.5.49.65: icmp
>         0x0000:  4500 1000 0f2d 0040 4001 e4c7 0a05 317e
>
> 3rd: 12:21:11.062900 IP 10.5.49.126 > 10.5.49.65: ICMP echo request, id
> 46597, seq 2, length 4076
>         0x0000:  4500 1000 0f2d 4000 4001 a507 0a05 317e
>
> 4th: 12:21:12.072915 IP 10.5.49.126 > 10.5.49.65: icmp
>         0x0000:  4500 1000 0f2d 0040 4001 e4c7 0a05 317e
>
> 5th: 12:21:13.082900 IP 10.5.49.126 > 10.5.49.65: ICMP echo request, id
> 46597, seq 4, length 4076
>         0x0000:  4500 1000 0f2d 4000 4001 a507 0a05 317e
>
> When payload is 4076, outgoing _every_ outgoing request begins with
>     4500 xxxx xxxx 4000 4001,
> while above we see every second request starting with
>     4500 xxxx xxxx 0040 4001.

> Does anybody know header composing off pat? Haven't seen these bits for
> quiet some time...

Glebius was recently in the networking stack in -HEAD, fiddling with
how IP headers are treated. Maybe this is some corner case fallout
from that?



adrian