On 20/08/2013 02:12, John-Mark Gurney wrote:
Joe Holden wrote this message on Mon, Aug 19, 2013 at 09:32 +0100:
Hm, I hadn't considered that... how do other OSes and vendors handle
this? (eg Linux?)
Just changing the connected route would probably suffice, or maybe if
any routes not added with default interface mtu could/are be flagged,
then those could be changed or not depending on what added them?
How do you know which routes are which? I believe that FreeBSD will
automaticly reduce the MTU if you decrease it, but it won't increase
it.. How do you know the difference between someone increasing the MTU
on the interface to allow a specific host to talk at the larger MTU and
wanting the rest of the hosts to talk at the larger MTU...
connected route is the one that should be changed, not others as you are
right in that regard, there may be routes that shouldn't be changed like
the default gateway or other hosts. Make it a tunable exposed via sysctl
or something, job done.
At a previous work place, we used this feature so that we could use
MTU 9k to other FreeBSD boxes to get better NFS performance, and but
keep the other windows boxes which didn't have MTU 9k compatible
interfaces talking on the same LAN...
vlan interfaces achieve the same thing without having to mess about with
mtus on routes and also give you an interface to work with, a much nicer
method comparatively.
Perhaps need someone with more experience of the network stack to wade
in here...
How difficult would it be to have ifconfig do it? As in, when MTU is
changed on an interface, if there is a prefix configured, update the MTU
as an rtsock message?
Cheers,
Joe
On 19/08/2013 08:00, Julian Elischer wrote:
The following reply was made to PR kern/181388; it has been noted by GNATS.
From: Julian Elischer <jul...@elischer.org>
To: bug-follo...@freebsd.org, j...@rewt.org.uk
Cc:
Subject: Re: kern/181388: [route] Routes not updated on mtu change
Date: Mon, 19 Aug 2013 14:57:22 +0800
The problem is that this is not as simple as it seems.
The route MTU MIGHT have been set by something other than the
interface MTU
in the first place.
The interface MTU is a default for the route MTU but is not the only
source.
This actuall bit me a couple of days ago when I was wonderign why my
interface was not sending 9K packets.. turns out you need to do
'ifconfig_xn0="DHCP mtu 9000"' in order to have your dncp
configured interface routes have the right size.
so, I'm agreeing with you , but noticing that there are complications.
_______________________________________________
freebsd-net@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-net
To unsubscribe, send any mail to "freebsd-net-unsubscr...@freebsd.org"
