Re: [PATCH v2 3/4] cpu_map: Add POWER11 cpu model support

On Tue, Mar 04, 2025 at 17:40:59 +0100, Peter Krempa wrote: > On Tue, Mar 04, 2025 at 10:40:23 -0500, Narayana Murty N wrote: >> Add POWER11 as a supported cpu model for ppc64. >> >> Signed-off-by: Narayana Murty N <nnmlinux(a)linux.ibm.com&gt; >> --- >> src/cpu_map/index.xml | 1 + >> src/cpu_map/meson.build | 1 + >> src/cpu_map/ppc64_POWER11.xml | 6 ++++++ >> tests/domaincapsdata/qemu_9.2.0.ppc64.xml | 1 + >> tests/qemucapabilitiesdata/caps_9.2.0_ppc64.xml | 4 ++-- >> ...memory-hotplug-ppc64-nonuma.ppc64-latest.abi-update.args | 2 +- >> .../memory-hotplug-ppc64-nonuma.ppc64-latest.abi-update.xml | 2 +- >> .../memory-hotplug-ppc64-nonuma.ppc64-latest.args | 2 +- >> .../memory-hotplug-ppc64-nonuma.ppc64-latest.xml | 2 +- >> .../ppc64-default-cpu-kvm-pseries-3.1.ppc64-latest.args | 2 +- >> .../ppc64-default-cpu-kvm-pseries-3.1.ppc64-latest.xml | 2 +- >> .../ppc64-default-cpu-kvm-pseries-4.2.ppc64-latest.args | 2 +- >> .../ppc64-default-cpu-kvm-pseries-4.2.ppc64-latest.xml | 2 +- >> tests/qemuxmlconfdata/pseries-cpu-le.ppc64-latest.args | 2 +- >> tests/qemuxmlconfdata/pseries-cpu-le.ppc64-latest.xml | 2 +- >> 15 files changed, 21 insertions(+), 12 deletions(-) >> create mode 100644 src/cpu_map/ppc64_POWER11.xml > [...] > >> diff --git a/tests/qemuxmlconfdata/memory-hotplug-ppc64-nonuma.ppc64-latest.abi-update.args b/tests/qemuxmlconfdata/memory-hotplug-ppc64-nonuma.ppc64-latest.abi-update.args >> index 4691595916..d163bdc378 100644 >> --- a/tests/qemuxmlconfdata/memory-hotplug-ppc64-nonuma.ppc64-latest.abi-update.args >> +++ b/tests/qemuxmlconfdata/memory-hotplug-ppc64-nonuma.ppc64-latest.abi-update.args >> @@ -12,7 +12,7 @@ XDG_CONFIG_HOME=/var/lib/libvirt/qemu/domain--1-QEMUGuest1/.config \ >> -object '{"qom-type":"secret","id":"masterKey0","format":"raw","file":"/var/lib/libvirt/qemu/domain--1-QEMUGuest1/master-key.aes"}' \ >> -machine pseries,usb=off,dump-guest-core=off \ >> -accel kvm \ >> --cpu power11 \ >> +-cpu POWER11 \ >> -m size=1310720k,slots=16,maxmem=4194304k \ >> -overcommit mem-lock=off \ >> -smp 1,sockets=1,cores=1,threads=1 \ >> diff --git a/tests/qemuxmlconfdata/memory-hotplug-ppc64-nonuma.ppc64-latest.abi-update.xml b/tests/qemuxmlconfdata/memory-hotplug-ppc64-nonuma.ppc64-latest.abi-update.xml >> index 886cfe0316..104409a46e 100644 >> --- a/tests/qemuxmlconfdata/memory-hotplug-ppc64-nonuma.ppc64-latest.abi-update.xml >> +++ b/tests/qemuxmlconfdata/memory-hotplug-ppc64-nonuma.ppc64-latest.abi-update.xml >> @@ -10,7 +10,7 @@ >> <boot dev='hd'/> >> </os> >> <cpu mode='custom' match='exact' check='none'> >> - <model fallback='forbid'>power11</model> >> + <model fallback='forbid'>POWER11</model> > This looks suspicious. qemu seems to report the lowercase version: > > $ qemu-system-ppc64 -cpu ? | grep -i power11 > power11_v2.0 PVR 00821200 > power11 (alias for power11_v2.0) > > Any reason you're defining it uppercase? Ideally put the explanation > into the commit message. Most likely because historically QEMU used upper case names and when they switched to lower case we had to implement a backward compatible translation. But I think we shouldn't do this for newly introduced models and just use lower case. Jirka