the bug of free_memmap

[Catalin Marinas]

On 4 January 2012 14:05, Hu Ruihuan <specter118 at gmail.com> wrote:
>    I have fount a bug in free_memmap function. When using sparsemem,
> and just a memory section is in a hole, the start_pfn parameter of
> free_memmap will be the last page of the section forward. Then the
> start_pg will be the address of the last page of the section forward
> plus the size of struct page, not the right one.

I don't fully get what the problem is. Is this with the latest kernel?
Could you please provide an example of your memory banks layout?

-- 
Catalin