On 9/1/25 16:24, Daniel P. Berrangé via Devel wrote:
> On Thu, Aug 28, 2025 at 10:29:41AM +0200, Claudio Fontana wrote:
>> it is very useful for platforms to know how much time a node is
>> spending running guests.
>>
>> Signed-off-by: Claudio Fontana <cfontana@xxxxxxx>
>> ---
>> include/libvirt/libvirt-host.h | 10 ++++++++
>> src/test/test_driver.c | 9 +++++--
>> src/util/virhostcpu.c | 5 +++-
>> tests/virhostcpudata/linux-cpustat-24cpu.out | 25 ++++++++++++++++++++
>> tools/virsh-host.c | 3 +++
>> 5 files changed, 49 insertions(+), 3 deletions(-)
>>
>> diff --git a/include/libvirt/libvirt-host.h b/include/libvirt/libvirt-host.h
>> index 8922d00e0c..2b899144d3 100644
>> --- a/include/libvirt/libvirt-host.h
>> +++ b/include/libvirt/libvirt-host.h
>> @@ -267,6 +267,16 @@ typedef enum {
>> */
>> # define VIR_NODE_CPU_STATS_INTR "intr"
>>
>> +/**
>> + * VIR_NODE_CPU_STATS_GUEST:
>> + *
>> + * The cumulative CPU time spent running guests,
>> + * since the node booting up (in nanoseconds).
>
> FYI, I think this should be qualified slight to say
>
> 'The cumulative CPU time spent running guest virtual CPUs,'
>
> because it does NOT include the overhead from host side
> emulation outside of vCPUs, which is often considered to
> be part of the time spent running guests.
>
>
> With regards,
> Daniel
Hello, I just sent a follow-up patch to clarify this:
[PATCH] libvirt-host: VIR_NODE_CPU_STATS_GUEST: clarify "guest" time
Thanks,
Claudio
