Ok, I've shaken the shared page table code hard, and fixed all the bugs I
could find.  It now runs reliably on all the configurations I have access
to, including regression runs of LTP.

I've audited the usage of page_table_lock, and I believe I have covered all
the places that also need to hold the pte_page_lock.

For reference, one of the tests was TPC-H.  My code reduced the number of
allocated pte_chains from 5 million to 50 thousand.

Dave McCracken

======================================================================
Dave McCracken          IBM Linux Base Kernel Team      1-512-838-3059
dmccr@us.ibm.com                                        T/L   678-3059