[PATCH] reduce rcu_head size [0/2]

[Dipankar Sarma <dipankar@in.ibm.com>]

This was originally proposed by Rusty. I have it in two patches
for a reason - rcu-no-arg changes the call_rcu() api and if it
is too late in 2.6 to introduce it, we could still do some
shrinking by applying the singly-linked-rcu patch. Other than
the documented changes, there is no subtle semantics change -
rcus are still invoked in the same order.
Andrew, this will probably break manfred's patches, but
512-cpu scalability can probably wait a little until
I get around to analyze those (my next task in hand) :)

I have tested the patches lightly on an x86 box.

Thanks
Dipankar