Re: [pve-devel] [PATCH manager 00/10] enable qemu vm architecture selection

[Fiona Ebner <f.ebner@proxmox.com>]

Am 28.01.26 um 1:29 PM schrieb Dominik Csapak:
> This series brings the emulated aarch64 vms to the gui. This could
> be configured on the api only previously.
> 
> The code also handles if the host has a non-x86 architecture, and is
> generally structured in a way that should make it (relatively) easy
> to add other architectures in the future.
> 
> Some fields/panels needed adaptions to hide some values or change some
> defaults, since not every combination that can be configured makes
> sense. (e.g. seabios on aarch64 cannot work currently)
> 
> Generally the backend allows many combinations that don't make sense on
> all architectures, so i tried to limit the users here to the 'correct'
> ones, but we should think about limiting them in the backend too.
> 
> Also, the backend made some inconsistent choices with different
> architectures, such as the default controller changes with the
> architecture, but things like 'bios' or 'scsihw' do not. So I worked
> around these things a bit differently.
> 
> What is still missing:
> * make the disk controllers selective per architecture
>   (e.g. aarch64 has no ide controller, ovmf+aarch64 cannot boot from sata)

Related, but maybe independent patch-wise:

* do not add the ISO(s) from the wizard as IDE.

> * cpumodel + cpuflags need to be architecture aware

For reference:
https://lore.proxmox.com/pve-devel/20260129131021.118199-1-f.ebner@proxmox.com/
(v2 because I had forgotten to adapt the endpoint for flags)

> 
> other nice to haves not yet done, that also invovle changing the backend:
> * 'virt' machine support in our api, so one can choose older versions

AFAICT, this is already supported by the API?

> * allow 'pcie' passthrough on 'virt'
Nice work so far! :)