Here is the code to do the actual relocation.  I believe there is a
race in this version of the code.  If the relocation code sleeps while
allocating memory and the rcu code frees the old page tables before
the tlbflush, then if another cpu is running the process in user
space, the page tables could be corrupted.  Not really a big deal, but
it needs to be fixed.  Probably by not scheduling the rcu free until
after all the page tables have been relocated.

    Ross