Re: [PATCH v5 14/15] s390x/pci: Implement migration for emulated devices

On 260805 17:02, Farhan Ali wrote:
> 
> On 8/3/2026 7:44 PM, Konstantin Shkolnyy wrote:
>> Implement zPCI device state migration, consequently enabling migration
>> of VMs that have emulated PCI devices, whether virtio or not.
>> Migration is allowed for devices whose function handle has the
>> FH_SHM_EMUL bit set. For these devices QEMU will save and restore the
>> state of its zPCI emulator.
>>
>> This will enable emulated PCI migration starting with s390-ccw- 
>> virtio-11.1.
>>
>> Passthrough devices will continue to block migration.
>>
>> Signed-off-by: Konstantin Shkolnyy<[email protected]>
>> ---
>>   hw/s390x/s390-pci-bus.c         | 189 +++++++++++++++++++++++++++++++-
>>   hw/s390x/s390-pci-inst.c        |   2 +-
>>   hw/s390x/s390-virtio-ccw.c      |   4 +
>>   include/hw/s390x/s390-pci-bus.h |   4 +
>>   4 files changed, 193 insertions(+), 6 deletions(-)
>>
>> diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c
>> index a94e24a2ac..9441c80ef4 100644
>> --- a/hw/s390x/s390-pci-bus.c
>> +++ b/hw/s390x/s390-pci-bus.c
>> @@ -26,6 +26,7 @@
>>   #include "hw/pci/pci_bridge.h"
>>   #include "hw/pci/msi.h"
>>   #include "exec/cpu-common.h"
>> +#include "migration/blocker.h"
>>   #include "qemu/error-report.h"
>>   #include "qemu/module.h"
>>   #include "system/physmem.h"
>> @@ -34,6 +35,11 @@
>>   #include "trace.h"
>> +static const Property phb_props[] = {
>> +    DEFINE_PROP_BOOL("x-zpci-emul-dev-migr-enabled", S390pciState,
>> +                     emul_dev_migr_enabled, true),
>> +};
>> +
>>   S390pciState *s390_get_phb(void)
>>   {
>>       static S390pciState *phb;
>> @@ -917,6 +923,23 @@ static void set_pbdev_info(S390PCIBusDevice *pbdev)
>>       pbdev->pci_group = s390_group_find(ZPCI_DEFAULT_FN_GRP);
>>   }
>> +static int s390_set_emul_dev_migration_blocker(S390pciState *s, Error 
>> **errp)
>> +{
>> +    if (s->emul_dev_migr_enabled) {
>> +        return 0;
>> +    }
>> +    error_setg(&s->emul_dev_migr_blocker,
>> +               "Migration disabled for emulated zPCI devices on this 
>> machine type");
>> +    return migrate_add_blocker(&s->emul_dev_migr_blocker, errp);
>> +}
>> +
> I was trying to test this patch series, and something i tried was doing 
> was a save/restore of a VM with just CCW devices. But with this patch 
> the save/restore of the VM fails with error:
> 
> error: Requested operation is not valid: cannot migrate domain: 
> Migration disabled for emulated zPCI devices on this machine type
> 
> I think this is a regression, as the save/restore works with the same 
> guest definition with current master branch. I also think this may break 
> live migration for guest with just CCW devices. Something to also note, 
> by default libvirt auto adds a pci-root controller to the guest 
> definition [1]. So i wonder if that's tripping up the migration blocker?
> 
> 
> [1] https://docs.redhat.com/en/documentation/red_hat_enterprise_linux/7/ 
> html/virtualization_deployment_and_administration_guide/sect- 
> guest_virtual_machine_device_configuration-configuring_device_controllers

Yes, indeed there is this device in qtree, even without any PCI devices 
configured in the libvirt XML, and it blocks migration:

       dev: s390-pcihost, id ""
         x-zpci-emul-dev-migr-enabled = false
         x-config-reg-migration-enabled = true
         bypass-iommu = false
         bus: s390-pcibus.0
           type s390-pcibus
         bus: pci.0
           type PCI

Instead of creating a single migration blocker for this s390-pcihost, 
I'll create one for each PCI device if this x-zpci-emul-dev-migr-enabled 
flag if false.