Re: [patch v3, kernel version 3.2.1] Source mode for macvlan interface

[Eric Dumazet <eric.dumazet@gmail.com>]

Le mercredi 25 janvier 2012 à 07:22 +0100, Štefan Gula a écrit :

> It's a result of need to have enough space allocated for sk_buff
> struct before called macvlan_fill_info. I know that it looks somehow
> messy. BUt before the macvlan_fill_info is called there is only one
> procedure calcit. The problem of calcit is that it doesn't have access
> to struct net_device *dev int it. It has only access to struct sk_buff
> *skb, so from that point it cannot be called directly (whole parsing
> code had to be copied there to get the *dev a calling
> if_nlmsg_size(dev) from it. On the other hand calcit returns global
> static value of min_ifinfo_dump_size and that is shared among all
> interfaces and modifying like this do the job properly. There was
> another different way and that was creating and registering a whole
> new PF_MACLVAN netlink group resulting in copying almost 95% percent
> of given code and modified it somehow to do the same.
> 
> If it should be result of separate commitment than ok, but it was
> result apparently of need only for use of macvlan that's why it was
> included in my patch.
> 
> So how should I proceed with it?

A separate patch for this part, as David said ?

The intent is to catch people attention on a particular point, instead
of unnoticed code in some huge unrelated patch.