On Fri, 2009-06-19 at 22:36 +0530, Sachin Sant wrote:

The offending commit seems to be
c868d550115b9ccc0027c67265b9520790f05601.
mm: Move pgtable_cache_init() earlier

If i revert this commit, the machine boots fine.

That is strange. If I revert that commit, I get breakages on machines
here. It would be interesting to understand what the problem is here,
as we -do- use that kmem cache for allocating page tables, so we do
need it initialized that early. (IE, we can't allow vmalloc for example
to be called before the page table caches are initialized).

This will need more debugging and understanding as to why it hangs.

Cheers,
Ben.