Re: [PATCH net-next v3 3/4] net: use skb->priority for overloading skb->dropcount and skb->reserved_tailroom instead of skb->mark

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

On Mon, 2015-02-23 at 20:56 +0200, Shmulik Ladkani wrote:
> Hi,
> 
> On Mon, 23 Feb 2015 19:52:03 +0200 Eyal Birger <eyal.birger@gmail.com> wrote:
> > Commit 977750076d98 ("af_packet: add interframe drop cmsg (v6)")
> > unionized skb->mark and skb->dropcount in order to allow recording
> > of the socket drop count while maintaining struct sk_buff size.
> > 
> > In order to allow for the skb->mark to be fetched by user-space code
> > it needs to be detached from this union; skb->priority is used
> > instead.
> > 
> > The purpose of overloading skb->priority is solely for retaining
> > struct sk_buff size; skb->priority is not used after the skb is
> > queued to the socket.
> > 
> > Use of aliased fields, skb->dropcount and skb->reserved_tailroom,
> > is either done after the skb is cloned (e.g. in packet_rcv()) or
> > before skb->priority is used (e.g. in IGMPv3/MLD/TCP).
> 
> Preserving a small skb size is an important cause.
> 
> In this case however, it seems that aliasing 'priority' sacrifices
> maintainability: 'priority' does not seem "naturally orthogonal" to
> 'drop_count' or 'reserved_tailroom'.
> 
> One might arm 'priority' in various code flows (future and existing),
> which may accidentally clash with the aliases. It is unexpected, hard to
> remember (especially since 'priority' wasn't formerly aliased), and
> sometimes it's simply not trivial to assure there's no clash.
> 
> May I suggest to unalias 'mark' out of the union?

My suggestion would be to move dropcount in skb->cb[] instead.

mark & reserved_tailroom can stay in the union.