Herbert Xu wrote:

On Thu, Sep 25, 2008 at 11:42:15AM +0300, Timo Teräs wrote:

quoted

Maybe we should make af_key.c code instead cache one entry before
sending it to user space. We could then get rid of the last hack
inside xfrm core walking code.

Yes that should work.  But you want to keep cache the skb instead
of the state since the latter can die.

Exactly, keep the converted entry in skb, since we have to allocate
that anyway. And when there's no more entries coming from enumerator
the sequence number of final skb is adjusted to zero.

Not sure if I can finish it today, but I think I should have the
time by tomorrow. Unless you have time to do it before that, I'll
post a patch later this week. Will update the similar xfrm_policy
patch for that too, before sending it.

- Timo