RPI-B VM panic

[Alan Cox]

On 06/13/2014 07:23, Hans Petter Selasky wrote:
> On 06/13/14 13:14, Hans Petter Selasky wrote:
>> Hi,
>>
>> Some debugging reveals that the previous element has pindex == 0 aswell.
>> So it turns out there can be multiple pages having same pindex, and then
>> the "mpred" logic falls apart :-(
>>
>> --HPS
>>
>
> Hi,
>
> Added some random printfs to vm_xxx.c:
>
> One observation is that allocated pages are busy, and we are not
> waiting for then to become non-busy ?
>

By default, we mark a page as busy when we allocate it.  At the end of
vm_fault(_hold), we unbusy it after it is mapped.

> I cannot debug this further until coming Monday ... Patch suggestions
> are welcome!
>
> It happens approximately 1 of 2 times during boot.
>
> --HPS
>
>
>