On 2019/6/20 20:04, Igor Mammedov wrote:
> On Tue, 14 May 2019 04:18:14 -0700
> Dongjiu Geng <gengdongjiu@xxxxxxxxxx> wrote:
>
>> Support this feature since version 4.1, disable it by
>> default in the old version.
>>
>> Signed-off-by: Dongjiu Geng <gengdongjiu@xxxxxxxxxx>
>> ---
>> hw/arm/virt.c | 6 ++++++
>> include/hw/arm/virt.h | 1 +
>> 2 files changed, 7 insertions(+)
>>
>> diff --git a/hw/arm/virt.c b/hw/arm/virt.c
>> index 5331ab7..7bdd41b 100644
>> --- a/hw/arm/virt.c
>> +++ b/hw/arm/virt.c
>> @@ -2043,8 +2043,14 @@ DEFINE_VIRT_MACHINE_AS_LATEST(4, 1)
>>
>> static void virt_machine_4_0_options(MachineClass *mc)
>> {
>> + VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
>> +
>> virt_machine_4_1_options(mc);
>> compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
>> + /* Disable memory recovery feature for 4.0 as RAS support was
>> + * introduced with 4.1.
>> + */
>> + vmc->no_ras = true;
>
> So it would mean that the feature is enabled unconditionally for
> new machine types and consumes resources whether user needs it or not.
>
> In light of the race for leaner QEMU and faster startup times,
> it might be better to make RAS optional and make user explicitly
> enable it using a machine option.
I will add a machine option to make RAS optional, do you think we should enable or disable it by default? I think it is better if we enable it by default.
>
>
>> }
>> DEFINE_VIRT_MACHINE(4, 0)
>>
>> diff --git a/include/hw/arm/virt.h b/include/hw/arm/virt.h
>> index 4240709..7f1a033 100644
>> --- a/include/hw/arm/virt.h
>> +++ b/include/hw/arm/virt.h
>> @@ -104,6 +104,7 @@ typedef struct {
>> bool disallow_affinity_adjustment;
>> bool no_its;
>> bool no_pmu;
>> + bool no_ras;
>> bool claim_edge_triggered_timers;
>> bool smbios_old_sys_ver;
>> bool no_highmem_ecam;
>
> .
>
