[ILUG] Bizarre networking problem

[Gavin McCullagh]

Hi,

On Fri, 29 Feb 2008, Niall O Broin wrote:

> On 29 Feb 2008, at 15:55, Gavin McCullagh wrote:
>
>> It looked from what you sent me like the linux machine wasn't doing SACK
>> (an oldish kernel perhaps?) but the solaris 8 obviously does SACK and I
>> gather Tiger does too.
>
> No, the linux box is running 2.6.15, which while not new, isn't exactly
> old, and is rather newer than the S8 kernel.  I expect Tiger does do SACK
> although I didn't use it to initiate connections.

Sorry, I misread your off-list mail.  It was:

"Here's snoop's view (from s8B) of a successful syn - ack pair 

makaluisdnifms -> ucpustc TCP D=22 S=34805 Syn               Seq=969049027 Len=0 Win=8760 Options=<mss 1460>
ucpustc -> makaluisdnifms TCP D=34805 S=22 Syn Ack=969049028 Seq=2034371054 Len=0 Win=24820 Options=<mss 1460>

and here its view of the syns which never get an ack                                                                                         
                                                                                                                                             
sce600 -> ucpustc      TCP D=22 S=32841 Syn Seq=1060432362 Len=0 Win=24820 Options=<mss 1460,sack 16843778-2695681118|>
sce600 -> ucpustc      TCP D=22 S=32841 Syn Seq=1060432362 Len=0 Win=24820 Options=<mss 1460,sack 16843778-686016305|>"

so makaluisdnifms seems to have had SACK off which is presumably a solaris
machine?

Gavin