Re: [Qemu-devel] address order of virtio-mmio devices

[Laszlo Ersek]

On 01/30/15 10:54, Daniel P. Berrange wrote:
> On Thu, Jan 29, 2015 at 08:05:50PM +0000, Peter Maydell wrote:
>> On 29 January 2015 at 19:47, Laszlo Ersek <address@hidden> wrote:
>>> On 01/29/15 20:12, Laszlo Ersek wrote:
>>>> If the guest kernel changed its "assignment strategy" at some point, but
>>>> earlier it used to match the comment (and the code), then whichever way
>>>> we shape the comment will be wrong for the other kernel strategy. :) So,
>>>> in that case this code is probably best left undisturbed.
>>>>
>>>> I'll try to dig out some kernel commit for more evidence.
>>>
>>> http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=70161ff3
>>
>> Thanks for digging that up -- I was half-thinking we might
>> just have got it wrong two years ago when we wrote the code :-)
>>
>> We should probably update the comment to say (a) what we were
>> trying to do (b) that we don't want to change it now because
>> it would break existing setups.
> 
> While it is clear there is no solution that works correctly with all
> kernels, I hate to think that we're going to stick with an ordering
> that is clearly wrong for modern kernels, forever going forward. The
> aarch64 world is only just starting out, so on balance I think we
> should optimize for the future rather than the past, since that gives
> right behaviour for orders of magnitude more people in the long term.
> 
> Also can we start using a versioned machine type for ARM, and make the
> new machine type have the correct ordering for current kernels.

Wei recently posted a patch that introduced versioning for machvirt. (I
didn't see the patch, only know about it.) If Peter agrees, I guess both
Wei's patch and mine could be applied. (Although, mine should be
reworked so that it affect only the new machtype.)

Thanks
Laszlo