On Fri, 12 Jul 2002, David Laight wrote:

> > I should probably mention I'm setting TCP_NODELAY to one on this, so I'd
> > expect any such heuristics to be deactivated. :-)
> >
> > Also, setting 'sysctl -w net.inet.tcp.ack_on_push=1' makes the problem go
> > away, but at an increased packet flow.
>
> Mmmm... the 'interwork with M$ stack' feature :-)
>
> Obviously a timestamped line trace would be useful.

See http://mail-index.netbsd.org/tech-net/2002/05/02/0004.html when I
first mentioend this.

> Still guessing....
> If the 'ack_on_push' flag makes a difference, then clearly (?)
> something is waiting for the data to be acked.

Yes, it's the kernel. I however believe it shouldn't be. :-)

> I wonder if the 'zero copy tcp' stuff is getting in the way
> - somehow stalling until TCP releases the 'borrowed' buffer.
> 4k is (probably) the system page size....

Nope. 1) that's not turned on, and 2) I've seen this well before Jason
made that change.

I'm wondering if some of the code thinks the window size is in bytes, not
integers.

Take care,

Bill