Re: [PATCH 1/4] flow: virtualize flow cache entry methods

["Timo Teräs" <timo.teras@iki.fi>]

Herbert Xu wrote:
> On Sun, Apr 04, 2010 at 01:50:16PM +0300, Timo Teräs wrote:
>> Because flow_cache_entry is per-cpu, and multiple entries (due to
>> different flows matching same policies, or same flow having multiple
>> per-cpu entries) can point to same policy. If we cached "dummy" objects
>> for even policies, then this would be better approach.
> 
> Oh yes of course.
> 
> But what we could do is embed most of flow_cache_entry into
> xfrm_policy (and xdst in your latter patches) along with the
> ops pointer.
> 
> Like this:
> 
> struct flow_cache_object {
> 	u16			family;
> 	u8			dir;
> 	u32			genid;
> 	struct flowi		key;
> 	struct flow_cache_ops **ops;
> };
> 
> struct flow_cache_entry {
> 	struct flow_cache_entry	*next;
> 	struct flow_cache_object *obj;
> };
> 
> struct xfrm_policy {
> 	struct flow_cache_object flo;
> 	...
> };
> 
> What do you think?

It would still not work for policies. For every policy X we
can get N+1 different matches with separate struct flowi contents.
It's not possible to put single struct flowi or any other of
the flow details in to xfrm_policy. It's a N-to-1 mapping. Not
a 1-to-1 mapping.