more strict KPI for vm_pager_get_pages()

[Alan Cox]

On 05/06/2015 06:45, Gleb Smirnoff wrote:
>   Hi!
>
>   I'm splitting the patch into a serie. This is step 1:
>
> Pagers are responsible to update the array of pages in
> case if they replace pages in an object. All array entries
> must be valid, if pager returns VM_PAGER_OK.
>
> Note: the only pager that replaces pages is sg_pager, and
> it does that correctly.
>

No, look again, e.g., device_pager.