[pve-devel] [PATCH v9 qemu-server 5/6] fix #2318: allow phys-bits and host-phys-bits CPU settings

[Fabian Ebner]

On 26.03.20 16:13, Stefan Reiter wrote:
> Can be specified for a particular VM or via a custom CPU model (VM takes
> precedence).
> 
> QEMU's default limit only allows up to 1TB of RAM per VM. Increasing the
> physical address bits available to a VM can fix this.
> 
> Signed-off-by: Stefan Reiter <s.reiter at proxmox.com>
> ---
>   PVE/QemuServer/CPUConfig.pm | 24 ++++++++++++++++++++++++
>   1 file changed, 24 insertions(+)
> 
> diff --git a/PVE/QemuServer/CPUConfig.pm b/PVE/QemuServer/CPUConfig.pm
> index fa09f4b..2b2529d 100644
> --- a/PVE/QemuServer/CPUConfig.pm
> +++ b/PVE/QemuServer/CPUConfig.pm
> @@ -149,6 +149,19 @@ my $cpu_fmt = {
>   	pattern => qr/$cpu_flag_any_re(;$cpu_flag_any_re)*/,
>   	optional => 1,
>       },
> +    'phys-bits' => {
> +	type => 'integer',
> +	minimum => 8,
> +	maximum => 64,
> +	description => "The physical memory address bits that are reported to the guest OS. Should be smaller or equal to the host's.",
> +	optional => 1,
> +    },
> +    'host-phys-bits' => {
> +	type => 'boolean',
> +	default => 0,
> +	description => "Whether to report the host's physical memory address bits. Overrides 'phys-bits' when set.",

Is it better to die when both are set, so that a user gets informed 
about the clashing options?

> +	optional => 1,
> +    },
>   };
>   
>   # $cpu_fmt describes both the CPU config passed as part of a VM config, as well
> @@ -472,6 +485,17 @@ sub get_cpu_options {
>       $cpu_str .= resolve_cpu_flags($pve_flags, $hv_flags, $custom_cputype_flags,
>   			      $vm_flags, $pve_forced_flags);
>   
> +    my $phys_bits = '';
> +    foreach my $conf ($custom_cpu, $cpu) {
> +	next if !defined($conf);
> +	if ($conf->{'host-phys-bits'}) {
> +	    $phys_bits = ",host-phys-bits=true";
> +	} elsif ($conf->{'phys-bits'}) {
> +	    $phys_bits = ",phys-bits=$conf->{'phys-bits'}";
> +	}
> +    }
> +    $cpu_str .= $phys_bits;
> +
>       return ('-cpu', $cpu_str);
>   }
>   
>