[pve-devel] Bug in Promox's OVMF implementation - resolution is garbled on MacOS, unless you reset machine on each boot?

[Dominik Csapak]

On 3/26/20 7:50 PM, Victor Hooi wrote:
> Hi.

Hi,

> 
> Also - I just saw this earlier thread in the mailing list:
> 
> https://pve.proxmox.com/pipermail/pve-devel/2020-February/041713.html
> 
> Could this be related to the issue of MacOS VMs not saving the UEFI 
> resolution? What version of qemu-server has this patch?

could be, it is not applied yet, but to test if it has any influence,
create a new efidisk on a directory storage (and not on zfs/lvm/rbd/etc.)
(i pinged the list for my patch btw)


>     I can confirm that yes - I do have a efidisk set up. The MacOS on
>     Proxmox tutorial
>     <https://www.nicksherlock.com/2019/10/installing-macos-catalina-10-15-on-proxmox-6/> which
>     I followed actually talks explicitly about setting up the efidisk.
> 
>     The graphical glitch does appear to be some bug between OVMF and
>     MacOS specifically. I have asked upstream (link
>     <https://edk2.groups.io/g/discuss/message/165>) at the Tiancore
>     project in case they know anything.
> 
>     Nicholas has patched the OVMF firmware - details
>     <https://www.nicksherlock.com/2018/04/patch-ovmf-to-support-macos-in-proxmox-5-1/> -
>     but as far as I know, it's only to revert two small commits:
> 
>         Proxmox 5 and 6’s version of the OVMF firmware includes two
>         commits (2ac1730 and 147fd35) that are intended to mark the
>         pagetables as read-only during startup. This was first seen in
>         Proxmox 5.1. This conflicts with the OsxAptioFixDrv drivers in
>         Clover, which expect to be able to modify the pagetables to
>         remap memory:
> 
> 
>     Do you happen to know if there is a fix in Proxmox that could make
>     this work again, without needing to patch it?

Depends, do you know (or can link to any mailinglist/issue/etc) why
this was done upstream and needs to be reverted?

We generally do not want to deviate from our upstream much, especially
if the use case is not supported by us (and afaik macos guests aren't)
and it is not clear if the patches hurt our other users.

kind regards
Dominik