Re: [Qemu-devel] [RFC PATCH qemu 1/4] memory: Postpone flatview and dispatch tree building till all devices are added

[Alexey Kardashevskiy]

On 08/09/17 00:30, Peter Maydell wrote:
> On 7 September 2017 at 15:27, Alexey Kardashevskiy <address@hidden> wrote:
>> On 07/09/17 19:30, Peter Maydell wrote:
>>> What happens if something in a device realize function tries
>>> to do a read from an AddressSpace?
>>
>>
>> address_space_memory is created before that loop but yes, address spaces
>> for devices are not rendered and QEMU crashes, needs fixing.
> 
> I'd be OK if we defined this to be not permitted, as long as
> we're reasonably happy that we don't currently do it anywhere
> and the error if it does happen is fairly clear about what's
> gone wrong.


QEMU actually only crashes with 2/4 + 3/4, with just 1/4 QEMU gets to
unassigned_mem_accepts() silently. Not sure how to convert this to a clear
error message.



-- 
Alexey