On Mon, Jun 15, 2026 at 11:01:21AM +0200, Claudio Fontana wrote:
> Hello Peter,
>
> On 6/15/26 10:47, Peter Krempa wrote:
> > On Mon, Jun 15, 2026 at 10:34:54 +0200, Claudio Fontana wrote:
> >> Fix VIR_NODE_CPU_STATS_GUEST to also account for guest nice time,
> >> similarly to how VIR_NODE_CPU_STATS_USER also accounts for nice time.
> >
> > Can you elaborate a bit what the problem is ... ideally in the commit
> > message.
>
> Well it is there, the problem is that as I introduced this last year
> I did not account for guest nice time.
>
> Hopefully we can still fix it as maybe no-one is relying on this yet..
>
> >
> >>
> >> VIR_NODE_CPU_STATS_GUEST was introduced in: b5878727c714
> >> ("util: virHostCPUGetStatsLinux: support VIR_NODE_CPU_STATS_GUEST")
> >>
> >> Extend the tests artificially to capture this specific case.
>
> The trap was that the system on which the tests were likely generated did not
> include any guest_nice time,
> so tests seemed to pass. But now as I revisited the issue I noticed the
> missing guest_nice.
>
> >>
> >> Fixes: b5878727c714d813c820ad4a1b695fbbb5ffc84e
> >
> > v11.7.0-2-gb5878727c7
> >
> >> Signed-off-by: Claudio Fontana <[email protected]>
> >> ---
> >> src/util/virhostcpu.c | 2 +-
> >> tests/virhostcpudata/linux-cpustat-24cpu.out | 8 ++++----
> >> tests/virhostcpudata/linux-cpustat-24cpu.stat | 4 ++--
> >> 3 files changed, 7 insertions(+), 7 deletions(-)
> >>
> >> diff --git a/src/util/virhostcpu.c b/src/util/virhostcpu.c
> >> index 1db096ba24..f2b25e940b 100644
> >> --- a/src/util/virhostcpu.c
> >> +++ b/src/util/virhostcpu.c
> >> @@ -882,7 +882,7 @@ virHostCPUGetStatsLinux(FILE *procstat,
> >> return -1;
> >>
> >> if (virHostCPUStatsAssign(¶ms[4],
> >> VIR_NODE_CPU_STATS_GUEST,
> >> - guest * TICK_TO_NSEC) < 0)
> >> + (guest + guest_nice) *
> >> TICK_TO_NSEC) < 0)
> >> return -1;
> >> return 0;
> >> }
> >
> > Shouldn't we rather introduce a new field rather than modify what the
> > old one does?
>
> I hope it's early enough that not many users are relying on
> VIR_NODE_CPU_STATS_GUEST ,
> and accounting it together would be in line with what is done for the other
> stats.
Yes, this feels like simply a bug fix to me, rather than an entirely
new category of data to report.
With regards,
Daniel
--
|: https://berrange.com ~~ https://hachyderm.io/@berrange :|
|: https://libvirt.org ~~ https://entangle-photo.org :|
|: https://pixelfed.art/berrange ~~ https://fstop138.berrange.com :|
