Chasing some invalid accesses to .init zone, I found that free_init_pages() 
was properly freeing the pages but virtual was still usable.

A poisoning (memset(page, 0xcc, PAGE_SIZE)) was done but this is not reliable.

A new config option DEBUG_INITDATA is introduced to mark this initdata as not 
present at all so that buggy code can trigger a fault.

This option is not meant for production machines because it may split one or 
two huge page (2MB or 4MB) into small pages and thus slow down kernel a bit.

(After that we could map non possible cpu percpu data to the initial 
percpudata that is included in .init and discarded in free_initmem())

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>