Re: [patch] TCP/IP delacks disabled with MPI

Alan Cox (alan@lxorguk.ukuu.org.uk)
Wed, 2 Jun 1999 17:12:41 +0100 (BST)


> Really? It is dubius idea. If keepalives will be lost,
> link will fail. And they will be lost, if share queue with data.

It works very well. They are marked high priority and some can
be lost happily without problems.

> No, Alan. I am ready to give up, but it is very dubiuos.
> dev_queue_xmit is used for data. It cannot be used before
> link setup is finished, because there is no link yet.

Of course it can. I bring the interface up and there is a link at
the physical layer, but no valid IP path. The wire is up, the IP isnt at
that point.

> I am sorry, but if CBQ is attached to the device, these
> guys who implemented keepalives in this manner risk to drop
> their control data to a blackhole or to police them to death.

There is no other sane way to do it. It works and it works well. I know
about the potential problem, but until there is a clean dev_queue_xmit_head()
function then it will have to stay

> It is interesting also, how Cisco HDLC keepalives look in tcpdump 8)

Grin.

Alan

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/