Skip to contentSkip to conversation

s390x/pci: mark zpci devices as unmigratable

PATCH In review Cornelia Huck 3 participants 6 messages Feb 1, 2019

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 s390x/pci: mark zpci devices as unmigratable

We currently don't migrate any state for zpci devices, which are coupled with standard pci devices. This means funny things happen when we e.g. try to migrate with a virtio-pci device but the s390x- specific zpci state is not migrated (vfio-pci is not affected, as it is not migratable anyway.)

Until this is fixed, mark zpci devices as unmigratable.

---

This is just a stop-gap measure to give us time to implement the needed migration code properly.

---
 hw/s390x/s390-pci-bus.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
hw/s390x/s390-pci-bus.c +10 -0
index c96a7cba34..96c7c18f3f 100644
@@ -1253,6 +1253,15 @@ static Property s390_pci_device_properties[] = {
     DEFINE_PROP_END_OF_LIST(),
 };
 
+static const VMStateDescription s390_pci_device_vmstate = {
+    .name = TYPE_S390_PCI_DEVICE,
+    /*
+     * TODO: add state handling here, so migration works at least with
+     * emulated pci devices on s390x
+     */
+    .unmigratable = 1,
+};
+
 static void s390_pci_device_class_init(ObjectClass *klass, void *data)
 {
     DeviceClass *dc = DEVICE_CLASS(klass);
@@ -1263,6 +1272,7 @@ static void s390_pci_device_class_init(ObjectClass *klass, void *data)
     dc->bus_type = TYPE_S390_PCI_BUS;
     dc->realize = s390_pci_device_realize;
     dc->props = s390_pci_device_properties;
+    dc->vmsd = &s390_pci_device_vmstate;
 }
 
 static const TypeInfo s390_pci_device_info = {
Signature · 1 line
2.17.2
Reviewed-by on v1 1/1
#1 lore raw
Quoted 41 lines · Cornelia Huck

We currently don't migrate any state for zpci devices, which are coupled with standard pci devices. This means funny things happen when we e.g. try to migrate with a virtio-pci device but the s390x- specific zpci state is not migrated (vfio-pci is not affected, as it is not migratable anyway.)

Until this is fixed, mark zpci devices as unmigratable.

---

This is just a stop-gap measure to give us time to implement the needed migration code properly.

---
 hw/s390x/s390-pci-bus.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
hw/s390x/s390-pci-bus.c +9 -0
index c96a7cba34..96c7c18f3f 100644
@@ -1253,6 +1253,15 @@ static Property s390_pci_device_properties[] = {
     DEFINE_PROP_END_OF_LIST(),
 };
 
+static const VMStateDescription s390_pci_device_vmstate = {
+    .name = TYPE_S390_PCI_DEVICE,
+    /*
+     * TODO: add state handling here, so migration works at least with
+     * emulated pci devices on s390x
+     */
+    .unmigratable = 1,
+};
+
 static void s390_pci_device_class_init(ObjectClass *klass, void *data)
 {
     DeviceClass *dc = DEVICE_CLASS(klass);

hw/s390x/s390-pci-bus.c static void s390_pci_device_class_init(ObjectClass *klass, void *data) · line 1278

Quoted 7 lines
hw/s390x/s390-pci-bus.c +1 -0
@@ -1263,6 +1272,7 @@ static void s390_pci_device_class_init(ObjectClass *klass, void *data)
     dc->bus_type = TYPE_S390_PCI_BUS;
     dc->realize = s390_pci_device_realize;
     dc->props = s390_pci_device_properties;
+    dc->vmsd = &s390_pci_device_vmstate;
 }
 
 static const TypeInfo s390_pci_device_info = {

I guess this should be good enough (e.g. pci-bridge without a zpci device should migrate "itself" I assume).

Signature · 3 lines
Thanks,

David / dhildenb
#2 lore raw
Quoted 47 lines · On Fri, 1 Feb 2019 14:29:47 +0100 David Hildenbrand <[email protected]> wrote:
Quoted 46 lines · Cornelia Huck

We currently don't migrate any state for zpci devices, which are coupled with standard pci devices. This means funny things happen when we e.g. try to migrate with a virtio-pci device but the s390x- specific zpci state is not migrated (vfio-pci is not affected, as it is not migratable anyway.)

Until this is fixed, mark zpci devices as unmigratable.

---

This is just a stop-gap measure to give us time to implement the needed migration code properly.

---
 hw/s390x/s390-pci-bus.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
hw/s390x/s390-pci-bus.c +10 -0
index c96a7cba34..96c7c18f3f 100644
@@ -1253,6 +1253,15 @@ static Property s390_pci_device_properties[] = {
     DEFINE_PROP_END_OF_LIST(),
 };
 
+static const VMStateDescription s390_pci_device_vmstate = {
+    .name = TYPE_S390_PCI_DEVICE,
+    /*
+     * TODO: add state handling here, so migration works at least with
+     * emulated pci devices on s390x
+     */
+    .unmigratable = 1,
+};
+
 static void s390_pci_device_class_init(ObjectClass *klass, void *data)
 {
     DeviceClass *dc = DEVICE_CLASS(klass);
@@ -1263,6 +1272,7 @@ static void s390_pci_device_class_init(ObjectClass *klass, void *data)
     dc->bus_type = TYPE_S390_PCI_BUS;
     dc->realize = s390_pci_device_realize;
     dc->props = s390_pci_device_properties;
+    dc->vmsd = &s390_pci_device_vmstate;
 }
 
Quoted 5 lines

hw/s390x/s390-pci-bus.c static void s390_pci_device_class_init(ObjectClass *klass, void *data) · line 1278

Quoted 2 lines
hw/s390x/s390-pci-bus.c +0 -0
 static const TypeInfo s390_pci_device_info = {

I guess this should be good enough (e.g. pci-bridge without a zpci device should migrate "itself" I assume).

I don't think we need to do anything special for those.

We probably need to go through the whole zpci code and check where we might miss some state handling. The original use case for that support was vfio-pci, which probably explains why migration had been neglected...

However, unmigratable zpci devices should be an effective stopper until this has been fixed.

Quoted 3 lines

Thanks!

Reviewed-by on v1 1/1
#3 lore raw
In reply to Cornelia Huck
Quoted 6 lines · Cornelia Huck

We currently don't migrate any state for zpci devices, which are coupled with standard pci devices. This means funny things happen when we e.g. try to migrate with a virtio-pci device but the s390x- specific zpci state is not migrated (vfio-pci is not affected, as it is not migratable anyway.)

Quoted 8 lines

Until this is fixed, mark zpci devices as unmigratable.

---

This is just a stop-gap measure to give us time to implement the needed migration code properly.

I imagine this means we'll want a single migration handler that will take care of PCI and zPCI in one go, that way the data stays coupled?

Just throwing thoughts out there.

Quoted 27 lines
---
  hw/s390x/s390-pci-bus.c | 10 ++++++++++
  1 file changed, 10 insertions(+)
hw/s390x/s390-pci-bus.c +9 -0
index c96a7cba34..96c7c18f3f 100644
@@ -1253,6 +1253,15 @@ static Property s390_pci_device_properties[] = {
      DEFINE_PROP_END_OF_LIST(),
  };
 
+static const VMStateDescription s390_pci_device_vmstate = {
+    .name = TYPE_S390_PCI_DEVICE,
+    /*
+     * TODO: add state handling here, so migration works at least with
+     * emulated pci devices on s390x
+     */
+    .unmigratable = 1,
+};
+
  static void s390_pci_device_class_init(ObjectClass *klass, void *data)
  {
      DeviceClass *dc = DEVICE_CLASS(klass);

hw/s390x/s390-pci-bus.c static void s390_pci_device_class_init(ObjectClass *klass, void *data) · line 1278

Quoted 7 lines
hw/s390x/s390-pci-bus.c +1 -0
@@ -1263,6 +1272,7 @@ static void s390_pci_device_class_init(ObjectClass *klass, void *data)
      dc->bus_type = TYPE_S390_PCI_BUS;
      dc->realize = s390_pci_device_realize;
      dc->props = s390_pci_device_properties;
+    dc->vmsd = &s390_pci_device_vmstate;
  }
 
  static const TypeInfo s390_pci_device_info = {

Good band-aid patch.!

#4 lore raw
Quoted 13 lines · Collin Walling
Quoted 12 lines · Cornelia Huck

We currently don't migrate any state for zpci devices, which are coupled with standard pci devices. This means funny things happen when we e.g. try to migrate with a virtio-pci device but the s390x- specific zpci state is not migrated (vfio-pci is not affected, as it is not migratable anyway.)

Until this is fixed, mark zpci devices as unmigratable.

---

Quoted 7 lines
Quoted 2 lines

This is just a stop-gap measure to give us time to implement the needed migration code properly.

I imagine this means we'll want a single migration handler that will take care of PCI and zPCI in one go, that way the data stays coupled?

Just throwing thoughts out there.

Guess this won't really be possible. Each device is to migrate "itself". We'll then also have to see which parts of the host bridge requires migration. Most probably not that easy. Migrating timers and such. Taking care of resets...

As I am planning to end my journey to wondeful zpci land for now (but I might eventually look into it again in the future), I won't be looking into migration.

If IBM has some resources for that, very much appreciated.

Signature · 3 lines
Thanks,

David / dhildenb
#5 lore raw
Quoted 27 lines · On Fri, 1 Feb 2019 17:06:26 +0100 David Hildenbrand <[email protected]> wrote:
Quoted 13 lines · Collin Walling
Quoted 12 lines · Cornelia Huck

We currently don't migrate any state for zpci devices, which are coupled with standard pci devices. This means funny things happen when we e.g. try to migrate with a virtio-pci device but the s390x- specific zpci state is not migrated (vfio-pci is not affected, as it is not migratable anyway.)

Until this is fixed, mark zpci devices as unmigratable.

---

Quoted 7 lines
Quoted 2 lines

This is just a stop-gap measure to give us time to implement the needed migration code properly.

I imagine this means we'll want a single migration handler that will take care of PCI and zPCI in one go, that way the data stays coupled?

Just throwing thoughts out there.

Guess this won't really be possible. Each device is to migrate "itself". We'll then also have to see which parts of the host bridge requires migration. Most probably not that easy. Migrating timers and such. Taking care of resets...

Quoted 5 lines

As I am planning to end my journey to wondeful zpci land for now (but I might eventually look into it again in the future), I won't be looking into migration.

If IBM has some resources for that, very much appreciated.

And I'd be happy to queue such patches (just as I did right now for this patch :)