Re: [Qemu-devel] [RFC] virtio-mmio: implement modern (v2) personality (virtio-1)

[Laszlo Ersek <lersek@redhat.com>]

On 08/02/19 11:20, Peter Maydell wrote:
> On Fri, 2 Aug 2019 at 01:26, Laszlo Ersek <lersek@redhat.com> wrote:
>> But it's extra work, not entirely risk-free (regressions), and I can't
>> tell if someone out there still uses virtio-mmio (despite me thinking
>> that would be unreasonable). I wouldn't like to see more work sunk into
>> it either way :)
> 
> The main reasons I still see people using virtio-mmio for
> the 'virt' board are:
>  * still using old but working command lines
>  * newer setups that were put together working from older tutorials
>    that recommended virtio-mmio because they predated virtio-pci
>    support being widespread
>  * using older (eg distro) kernels -- for 32-bit kernels in
>    particular it was a while before the virtio-pci support
>    got built in the default configs I think, and then longer
>    again until those got into stable distro releases

There was one time when edk2 core modules gained a feature for marking
the DXE phase stack non-executable. We happily enabled it in OVMF. Then
people reported that UEFI grub in their old Debian guests had broken --
on their hosts carrying OVMF binaries built from upstream. :) We had to
flip the default off in OVMF, and we've stuck with that ever since...

  https://github.com/tianocore/edk2/commit/d20b06a3afdf

> I wouldn't be surprised if some of those applied also to
> via-OVMF boot setups as well as direct kernel boot. So it
> depends a bit what your tolerance for breaking existing
> user setups is.

Near zero. :)

Thanks
Laszlo