svn commit: r209220 - projects/ppc64/sys/powerpc/aim

[Alan Cox]

Nathan Whitehorn wrote:
> Author: nwhitehorn
> Date: Tue Jun 15 21:46:07 2010
> New Revision: 209220
> URL: http://svn.freebsd.org/changeset/base/209220
>
> Log:
>   Implement a page table spilling mechanism to allow the kernel to continue
>   when user-space applications have filled all the slots in a PTEG. This
>   relies on VM gracefully remapping pages when it gets faults on pages
>   it thinks are mapped already, which it seems to do in limited
>   forkbomb-based testing.
>   
>   Comments by alc appreciated before merging to HEAD.
>   

As long the mapping requested of pmap_enter() is valid upon return from 
that function, a pmap implementation is allowed to destroy any other 
mappings to pageable pages, i.e., pages and mappings that aren't wired, 
at any time.  So, based upon the above description, I don't think that 
you are doing anything inappropriate.

Regards,
Alan