> Followup to: <Pine.LNX.3.95.961103080752.266A-100000@mikeg.weiden.de>
> By author: "Michael L. Galbraith" <mikeg@mikeg.weiden.de>
> In newsgroup: linux.dev.kernel
> >
> > Hello Everyone,
> >
> > I tried to adjust the mtu on an interface with the result:
> > SIOCSIFMTU: Invalid argument
>
> Care to specify what kernel version you're using?
Redface. I noticed the omission right after xmit. 2.0.24 was my victim.
>
> > No interface (except lo) can be set to an mtu of over 1500 period.
> > Did I miss something?
> > Rules change on me? I used to be able to go up to 4096-header.
>
> No, a bug was fixed. 1500 is the maximum MTU for Ethernet.
Bug? I've set the mtu > 1500 on other systems too.
>
> > It got a lot more interesting when Mr. Speedtypist here missed the
> > "mtu" part. I inadvertently entered "ifconfig eth0 2000", with the
> > following result:
> >
> > Nov 3 07:53:53 mikeg kernel: Unable to handle kernel paging request
> > at virtual address f63a3336
>
> Now this is a pretty serious kernel bug. "ifconfig eth0 2000" should
> be interpreted as 0.0.7.208; I don't know why that barfed -- seems
> pretty serious. On my system (2.0.20) it worked flawlessly; with the
> expected result:
That's the point - not > 1500 mtu - that is just a curiosity.
Gated shouldn't be able to trash the kernel - even with my assistance.
I repeated this several times to make sure it wasn't a fluke. The
address is being interpreted properly, and _sometimes_ the machine stays
up long enough to ifconfig back to the correct address.
Mike