Re: Remains of seq_file conversion for DECnet, plus fixes

[Steven Whitehouse <steve@gw.chygwyn.com>]

Hi,

> 
>    From: Steven Whitehouse <steve@gw.chygwyn.com>
>    Date: Thu, 1 May 2003 14:27:08 +0100 (BST)
>    
>    Its got big again, so I've put it here, if you'd like it in bits rather than
>    all one lump, just shout:
> 
> Please do this because it's easier for me to evaluate small patches
> doing one thing at a time and you're also making modifications to
> things outside of decnet.
>
Ok. I'll do that later this afternoon.
 
> About update_pmtu().  You are not required to implement this.
> All of these places you see dereferencing dst->update_mtu()
> know that they have an ipv4/ipv6 route.  Or do you know some
> exception to this?
> 
I was thinking of that bit of code in ip_gre.c which I sent the fix for
recently. Unless I've missed something it calls update_pmtu on the dst
which is passed by the encapsulated protocol, although I've not actually
tested that.

Thats the reason I wasn't sure about the fix which I sent... it deleted the
duplicated call only for IPv4 rather than the one which appears to get called
for every protocol as I wasn't sure what was intended.

Either way it seems reasonable to pass this information back to the
protocol so perhaps it should say (in ipgre_tunnel_xmit()):

	if (skb->dst && skb->dst->ops->update_pmtu)
		skb->dst->ops->update_pmtu(skb->dst, mtu);
?

Steve.