OLPC "iperf hang" notes

[David Miller]

From: Mitch Bradley <wmb at firmworks.com>
Date: Thu, 01 Mar 2007 13:07:45 -1000

> I wonder what is causing the packet reordering?  Does iperf explicitly 
> force reordering in order to stress test the TCP?

no.

> I would expect out-of-order packets to be extremely rare - virtually
> nonexistent - within the confines of a single machine.

This is especially not true on SMP.

And for a wireless mesh I'd expect lots of reordering.

But all this time is wasted ether.  Add tracing to the device
driver on the transmit side and figure out where what it's
doing when the connection is hung like this, is it's transmit
queue state down?  Is the device in some error state?

Stop tracing the wire, and start adding debugging to the driver
to figure out what's going on exactly on the sender side.