| Message ID | 20190902120222.6179-1-imammedo@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | x86: do not advertise die-id in query-hotpluggbale-cpus if '-smp dies' is not set | expand |
On Mon, Sep 02, 2019 at 08:02:22AM -0400, Igor Mammedov wrote: > Commit 176d2cda0 (i386/cpu: Consolidate die-id validity in smp context) added > new 'die-id' topology property to CPUs and exposed it via QMP command > query-hotpluggable-cpus, which broke -device/device_add cpu-foo for existing > users that do not support die-id/dies yet. That's would be fine if it happened > to new machine type only but it also happened to old machine types, > which breaks migration from old QEMU to the new one, for example following CLI: > > OLD-QEMU -M pc-i440fx-4.0 -smp 1,max_cpus=2 \ > -device qemu64-x86_64-cpu,socket-id=1,core-id=0,thread-id > is not able to start with new QEMU, complaining about invalid die-id. > > After discovering regression, the patch > "pc: Don't make die-id mandatory unless necessary" > makes die-id optional so old CLI would work. > > However it's not enough as new QEMU still exposes die-id via query-hotpluggbale-cpus > QMP command, so the users that started old machine type on new QEMU, using all > properties (including die-id) received from QMP command (as required), won't be > able to start old QEMU using the same properties since it doesn't support die-id. > > Fix it by hiding die-id in query-hotpluggbale-cpus for all machine types in case > '-smp dies' is not provided on CLI or -smp dies = 1', in which case smp_dies == 1 > and APIC ID is calculated in default way (as it was before DIE support) so we won't > need compat code as in both cases the topology provided to guest via CPUID is the same. > > Signed-off-by: Igor Mammedov <imammedo@redhat.com> Reviewed-by: Eduardo Habkost <ehabkost@redhat.com> Queued on machine-next.
diff --git a/hw/i386/pc.c b/hw/i386/pc.c index cf06d8a83f..f802741061 100644 --- a/hw/i386/pc.c +++ b/hw/i386/pc.c @@ -2913,8 +2913,10 @@ static const CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *ms) ms->smp.threads, &topo); ms->possible_cpus->cpus[i].props.has_socket_id = true; ms->possible_cpus->cpus[i].props.socket_id = topo.pkg_id; - ms->possible_cpus->cpus[i].props.has_die_id = true; - ms->possible_cpus->cpus[i].props.die_id = topo.die_id; + if (pcms->smp_dies > 1) { + ms->possible_cpus->cpus[i].props.has_die_id = true; + ms->possible_cpus->cpus[i].props.die_id = topo.die_id; + } ms->possible_cpus->cpus[i].props.has_core_id = true; ms->possible_cpus->cpus[i].props.core_id = topo.core_id; ms->possible_cpus->cpus[i].props.has_thread_id = true;
Commit 176d2cda0 (i386/cpu: Consolidate die-id validity in smp context) added new 'die-id' topology property to CPUs and exposed it via QMP command query-hotpluggable-cpus, which broke -device/device_add cpu-foo for existing users that do not support die-id/dies yet. That's would be fine if it happened to new machine type only but it also happened to old machine types, which breaks migration from old QEMU to the new one, for example following CLI: OLD-QEMU -M pc-i440fx-4.0 -smp 1,max_cpus=2 \ -device qemu64-x86_64-cpu,socket-id=1,core-id=0,thread-id is not able to start with new QEMU, complaining about invalid die-id. After discovering regression, the patch "pc: Don't make die-id mandatory unless necessary" makes die-id optional so old CLI would work. However it's not enough as new QEMU still exposes die-id via query-hotpluggbale-cpus QMP command, so the users that started old machine type on new QEMU, using all properties (including die-id) received from QMP command (as required), won't be able to start old QEMU using the same properties since it doesn't support die-id. Fix it by hiding die-id in query-hotpluggbale-cpus for all machine types in case '-smp dies' is not provided on CLI or -smp dies = 1', in which case smp_dies == 1 and APIC ID is calculated in default way (as it was before DIE support) so we won't need compat code as in both cases the topology provided to guest via CPUID is the same. Signed-off-by: Igor Mammedov <imammedo@redhat.com> --- hw/i386/pc.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)