Ingo Molnar wrote:

nice one! I suspect we could simplify some of the less 
performance-critical open-coded pagetable walker loops in the kernel 
with this? (i'm not even sure it's all that much of a performance loss 
to pass a function pointer around - would be nice to convert say 
mprotect() to this and see the performance difference?)
  

apply_page_to_range has the side-effect of allocating all the pagetable
levels for the address range it walks.  Xen uses this to populate
pagetables (for example, build out the pagetable, and let the hypervisor
plug a mapping into the pte page).

If that's OK for the other uses, then it should be good.

    J