[PATCH v7 14/15] s390x/pci: Implement migration for emulated devices

[Konstantin Shkolnyy <kshk@linux.ibm.com>]

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-11.2.

Passthrough devices will continue to block migration.

Reviewed-by: Farhan Ali<alifm@linux.ibm.com>
Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
---
 hw/s390x/s390-pci-bus.c         | 195 +++++++++++++++++++++++++++++++-
 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, 199 insertions(+), 6 deletions(-)

diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c
index a94e24a2ac..4ac845328f 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;
@@ -1129,6 +1135,46 @@ static int s390_pci_interp_plug(S390pciState *s, S390PCIBusDevice *pbdev)
     return 0;
 }
 
+static int s390_set_emul_dev_migration_blocker(S390PCIBusDevice *pbdev,
+        S390pciState *s, Error **errp)
+{
+    if (s->emul_dev_migr_enabled) {
+        return 0;
+    }
+    error_setg(&pbdev->emul_dev_migr_blocker,
+               "Migration blocked on this machine type by emulated zPCI device "
+               "fh 0x%x uid %d fid %d", pbdev->fh, pbdev->uid, pbdev->fid);
+    return migrate_add_blocker(&pbdev->emul_dev_migr_blocker, errp);
+}
+
+static void s390_clear_emul_dev_migration_blocker(S390PCIBusDevice *pbdev)
+{
+    if (pbdev->emul_dev_migr_blocker) {
+        migrate_del_blocker(&pbdev->emul_dev_migr_blocker);
+    }
+}
+
+static int s390_set_passthrough_migration_blocker(S390PCIBusDevice *pbdev,
+                                            Error **errp)
+{
+    pbdev->passthrough_migr_blocker = NULL;
+
+    if (pbdev->fh & FH_SHM_EMUL) {
+        return 0;
+    }
+    error_setg(&pbdev->passthrough_migr_blocker,
+               "Migration blocked by passthrough zPCI device "
+               "fh 0x%x uid %d fid %d", pbdev->fh, pbdev->uid, pbdev->fid);
+    return migrate_add_blocker(&pbdev->passthrough_migr_blocker, errp);
+}
+
+static void s390_clear_passthrough_migration_blocker(S390PCIBusDevice *pbdev)
+{
+    if (pbdev->passthrough_migr_blocker) {
+        migrate_del_blocker(&pbdev->passthrough_migr_blocker);
+    }
+}
+
 static void s390_pcihost_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
                               Error **errp)
 {
@@ -1242,9 +1288,20 @@ static void s390_pcihost_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
             pbdev->rtr_avail = false;
         }
 
+        if (s390_set_passthrough_migration_blocker(pbdev, errp) != 0) {
+            return;
+        }
+
+        if (s390_set_emul_dev_migration_blocker(pbdev, s, errp) != 0) {
+            s390_clear_passthrough_migration_blocker(pbdev);
+            return;
+        }
+
         if (s390_pci_msix_init(pbdev) && !pbdev->interp) {
             error_setg(errp, "MSI-X support is mandatory "
                        "in the S390 architecture");
+            s390_clear_emul_dev_migration_blocker(pbdev);
+            s390_clear_passthrough_migration_blocker(pbdev);
             return;
         }
 
@@ -1284,6 +1341,9 @@ static void s390_pcihost_unplug(HotplugHandler *hotplug_dev, DeviceState *dev,
             return;
         }
 
+        s390_clear_emul_dev_migration_blocker(pbdev);
+        s390_clear_passthrough_migration_blocker(pbdev);
+
         s390_pci_generate_plug_event(HP_EVENT_STANDBY_TO_RESERVED,
                                      pbdev->fh, pbdev->fid);
         bus = pci_get_bus(pci_dev);
@@ -1451,6 +1511,7 @@ static void s390_pcihost_class_init(ObjectClass *klass, const void *data)
     hc->unplug_request = s390_pcihost_unplug_request;
     hc->unplug = s390_pcihost_unplug;
     msi_nonbroken = true;
+    device_class_set_props(dc, phb_props);
 }
 
 static const TypeInfo s390_pcihost_info = {
@@ -1464,10 +1525,33 @@ static const TypeInfo s390_pcihost_info = {
     }
 };
 
+/* Return a unique bus "path" for zpci device */
+static char *s390_pci_bus_get_dev_path(DeviceState *dev)
+{
+    S390PCIBusDevice *pbdev = S390_PCI_DEVICE(dev);
+    return g_strdup_printf("uid-%04x", pbdev->uid);
+}
+
+static void s390_pcibus_class_init(ObjectClass *oc, const void *data)
+{
+    BusClass *bc = BUS_CLASS(oc);
+    bc->get_dev_path = s390_pci_bus_get_dev_path;
+}
+
 static const TypeInfo s390_pcibus_info = {
     .name = TYPE_S390_PCI_BUS,
     .parent = TYPE_BUS,
     .instance_size = sizeof(S390PCIBus),
+    /*
+     * Implement get_dev_path() to provide each zpci device with a unique
+     * stable UID-based bus "path". The "path" is used as part of idstr in the
+     * migration stream, making idstr unique and instance_id always 0.
+     * For migration to succeed, (idstr+instance_id) must match those generated
+     * during QEMU start. Without unique idstr, QEMU will generate variable
+     * instance_id to distinquish devices, and that instance_id can change
+     * if a device is unplugged and plugged back, preventing migration.
+     */
+    .class_init = s390_pcibus_class_init,
 };
 
 static uint16_t s390_pci_generate_uid(S390pciState *s)
@@ -1613,13 +1697,114 @@ static const Property s390_pci_device_properties[] = {
                      true),
 };
 
-static const VMStateDescription s390_pci_device_vmstate = {
-    .name = TYPE_S390_PCI_DEVICE,
+static int s390_pci_device_pre_load(void *opaque)
+{
+    S390PCIBusDevice *pbdev = S390_PCI_DEVICE(opaque);
+    S390PCIBusDevice *found_pbdev;
+
+    /*
+     * Because state loading can change pbdev->idx make sure pbdev is removed
+     * from the table before that happens. The table type used stores a pointer
+     * to pbdev->idx and becomes corrupt if idx is changed from outside. But be
+     * careful to not remove instead another pbdev whose state might have been
+     * loaded earlier and that got assigned this idx value and had therefore
+     * already replaced our pbdev in the table. post_load() will reinsert our
+     * pbdev into the table.
+     */
+    found_pbdev = g_hash_table_lookup(s390_get_phb()->zpci_table, &pbdev->idx);
+    assert(found_pbdev);
+    if (found_pbdev == pbdev) {
+        g_hash_table_remove(s390_get_phb()->zpci_table, &pbdev->idx);
+    }
+
+    return 0;
+}
+
+static int s390_pci_device_post_load(void *opaque, int version_id)
+{
+    S390PCIBusDevice *pbdev = S390_PCI_DEVICE(opaque);
+
+    /*
+     * Now that pbdev->idx has been loaded, use it to place pbdev back into
+     * the table. This may replace a different not-yet-state-loaded pbdev,
+     * but pre_load() handles this case.
+     */
+    g_hash_table_replace(s390_get_phb()->zpci_table, &pbdev->idx, pbdev);
+
+    /*
+     * Regenerate IOMMU state, including IOTLB contents and QEMU memory regions.
+     */
+    if (pbdev->iommu_enabled) {
+        assert(pbdev->iommu);
+        if (s390_pci_is_translation_enabled(pbdev->g_iota)) {
+            s390_pci_iommu_enable(pbdev);
+            s390_pci_ioat_replay(pbdev);
+        } else {
+            s390_pci_iommu_direct_map_enable(pbdev);
+        }
+    }
+
     /*
-     * TODO: add state handling here, so migration works at least with
-     * emulated pci devices on s390x
+     * Guest sets fmb_addr by mpcifc.ZPCI_MOD_FC_SET_MEASURE instruction,
+     * whose handler consequently starts fmb_timer. We may need to restart it.
      */
-    .unmigratable = 1,
+    if (pbdev->fmb_addr) {
+        assert(!pbdev->fmb_timer);
+        assert(pbdev->pci_group);
+        pbdev->fmb_timer = timer_new_ms(QEMU_CLOCK_VIRTUAL,
+                                        fmb_update, pbdev);
+        timer_mod(pbdev->fmb_timer,
+                  qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) +
+                                    pbdev->pci_group->zpci_group.mui);
+    }
+    return 0;
+}
+
+static const VMStateDescription s390_pci_device_vmstate = {
+    .name = TYPE_S390_PCI_DEVICE,
+    .version_id = 1,
+    .minimum_version_id = 1,
+    .pre_load = s390_pci_device_pre_load,
+    .post_load = s390_pci_device_post_load,
+    .fields = (const VMStateField[]) {
+        VMSTATE_UINT32(state, S390PCIBusDevice),
+        VMSTATE_UINT16(uid, S390PCIBusDevice),
+        VMSTATE_UINT32(idx, S390PCIBusDevice),
+        VMSTATE_UINT32(fh, S390PCIBusDevice),
+        VMSTATE_UINT32(fid, S390PCIBusDevice),
+        VMSTATE_BOOL(fid_defined, S390PCIBusDevice),
+        VMSTATE_UINT64(fmb_addr, S390PCIBusDevice),
+        VMSTATE_UINT32(fmb.format, S390PCIBusDevice),
+        VMSTATE_UINT32(fmb.sample, S390PCIBusDevice),
+        VMSTATE_UINT64(fmb.last_update, S390PCIBusDevice),
+        VMSTATE_UINT64_ARRAY(fmb.counter, S390PCIBusDevice,
+                ARRAY_SIZE(((S390PCIBusDevice *)0)->fmb.counter)),
+        VMSTATE_UINT64(fmb.fmt0.dma_rbytes, S390PCIBusDevice),
+        VMSTATE_UINT64(fmb.fmt0.dma_wbytes, S390PCIBusDevice),
+        VMSTATE_UINT8(isc, S390PCIBusDevice),
+        VMSTATE_UINT16(noi, S390PCIBusDevice),
+        VMSTATE_UINT8(sum, S390PCIBusDevice),
+        VMSTATE_UINT8(pft, S390PCIBusDevice),
+        VMSTATE_UINT64(routes.adapter.ind_addr, S390PCIBusDevice),
+        VMSTATE_UINT64(routes.adapter.summary_addr, S390PCIBusDevice),
+        VMSTATE_UINT64(routes.adapter.ind_offset, S390PCIBusDevice),
+        VMSTATE_UINT32(routes.adapter.summary_offset, S390PCIBusDevice),
+        VMSTATE_UINT32(routes.adapter.adapter_id, S390PCIBusDevice),
+        VMSTATE_BOOL(iommu_enabled, S390PCIBusDevice),
+        VMSTATE_UINT64(g_iota, S390PCIBusDevice),
+        VMSTATE_UINT64(pba, S390PCIBusDevice),
+        VMSTATE_UINT64(pal, S390PCIBusDevice),
+        VMSTATE_UINT64(max_dma_limit, S390PCIBusDevice),
+        VMSTATE_PTR_TO_IND_ADDR(summary_ind, S390PCIBusDevice),
+        VMSTATE_PTR_TO_IND_ADDR(indicator, S390PCIBusDevice),
+        VMSTATE_BOOL(pci_unplug_request_processed, S390PCIBusDevice),
+        VMSTATE_BOOL(unplug_requested, S390PCIBusDevice),
+        VMSTATE_BOOL(interp, S390PCIBusDevice),
+        VMSTATE_BOOL(forwarding_assist, S390PCIBusDevice),
+        VMSTATE_BOOL(aif, S390PCIBusDevice),
+        VMSTATE_BOOL(rtr_avail, S390PCIBusDevice),
+        VMSTATE_END_OF_LIST()
+    }
 };
 
 static void s390_pci_device_class_init(ObjectClass *klass, const void *data)
diff --git a/hw/s390x/s390-pci-inst.c b/hw/s390x/s390-pci-inst.c
index f93db10c81..6b0742d143 100644
--- a/hw/s390x/s390-pci-inst.c
+++ b/hw/s390x/s390-pci-inst.c
@@ -1122,7 +1122,7 @@ static int fmb_do_update(S390PCIBusDevice *pbdev, int offset, uint64_t val,
     return ret;
 }
 
-static void fmb_update(void *opaque)
+void fmb_update(void *opaque)
 {
     S390PCIBusDevice *pbdev = opaque;
     int64_t t = qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL);
diff --git a/hw/s390x/s390-virtio-ccw.c b/hw/s390x/s390-virtio-ccw.c
index 06e5def909..b22c27b066 100644
--- a/hw/s390x/s390-virtio-ccw.c
+++ b/hw/s390x/s390-virtio-ccw.c
@@ -939,8 +939,12 @@ static void ccw_machine_11_1_instance_options(MachineState *machine)
 
 static void ccw_machine_11_1_class_options(MachineClass *mc)
 {
+    static GlobalProperty compat[] = {
+        { TYPE_S390_PCI_HOST_BRIDGE, "x-zpci-emul-dev-migr-enabled", "off" },
+    };
     ccw_machine_11_2_class_options(mc);
     compat_props_add(mc->compat_props, hw_compat_11_1, hw_compat_11_1_len);
+    compat_props_add(mc->compat_props, compat, G_N_ELEMENTS(compat));
 }
 DEFINE_CCW_MACHINE(11, 1);
 
diff --git a/include/hw/s390x/s390-pci-bus.h b/include/hw/s390x/s390-pci-bus.h
index 17ecf3e0da..66c1f04ead 100644
--- a/include/hw/s390x/s390-pci-bus.h
+++ b/include/hw/s390x/s390-pci-bus.h
@@ -340,6 +340,8 @@ struct S390PCIBusDevice {
     uint16_t uid;
     uint32_t idx;
     uint32_t fh;
+    Error *emul_dev_migr_blocker;
+    Error *passthrough_migr_blocker;
     uint32_t fid;
     bool fid_defined;
     uint64_t fmb_addr;
@@ -393,6 +395,7 @@ struct S390pciState {
     QTAILQ_HEAD(, S390PCIDMACount) zpci_dma_limit;
     QTAILQ_HEAD(, S390PCIGroup) zpci_groups;
     uint8_t next_sim_grp;
+    bool emul_dev_migr_enabled;
 };
 
 S390pciState *s390_get_phb(void);
@@ -418,5 +421,6 @@ S390PCIBusDevice *s390_pci_find_dev_by_pci(S390pciState *s,
 S390PCIBusDevice *s390_pci_find_next_avail_dev(S390pciState *s,
                                                S390PCIBusDevice *pbdev);
 void s390_pci_ism_reset(void);
+void fmb_update(void *opaque);
 
 #endif
-- 
2.34.1