This will eventually need to be fixed in kernel but for now to allow work to progress alter mab so it will not attempt to place a table across a page boundary.
If people could give this patch some proper testing it would be good as there may be some corner cases I missed constructing this. I have boot tested the result on Arndale.
While working on this I discovered that write_table was allocating buffers which were never used and never had free() called on them.
Thanks
Graeme