[RFC PATCH v2 7/8] pci: automatically unplug a PCI card before migration

[Laurent Vivier]

We have moved all the functions needed by failover to unplug a card to the
PCI subsystem.

A side effect of this change is we can implement automatic hotplug/unplug
of any PCI card during migration without using a failover virtio-net card.
For that, we need to introduce a new PCI device property,
"unplug-on-migration", we can set to "true" or "on" if we want QEMU unplugs
the card before the migration and plugs it back on the destination side
after the migration.

We modify the pci_dev_hide_device() function to check for the
"unplug-on-migration" property on the command line.
If it is present, the device is hidden on startup only on the destination
side and it will be unplugged before the migration.

To implement the "unplug-on-migration" property, we add a post_load
function in vmstate_pcibus to hotplug the card after the migration
(bus_post_load() and pci_dev_replug_on_migration()). This is not
needed with virtio-net failover because the device is plugged back
by the virtio-net device during the features migration
(VIRTIO_NET_F_STANDBY)

Signed-off-by: Laurent Vivier <lvivier@redhat.com>
---
 include/hw/pci/pci.h |  1 +
 hw/pci/pci.c         | 76 ++++++++++++++++++++++++++++++++++++++------
 hw/vfio/pci.c        |  2 +-
 3 files changed, 69 insertions(+), 10 deletions(-)

diff --git a/include/hw/pci/pci.h b/include/hw/pci/pci.h
index d35214144d1b..e02d965c064f 100644
--- a/include/hw/pci/pci.h
+++ b/include/hw/pci/pci.h
@@ -362,6 +362,7 @@ struct PCIDevice {
 
     /* ID of standby device in net_failover pair */
     char *failover_pair_id;
+    bool unplug_on_migration;
     Notifier migration_state;
     uint32_t acpi_index;
 };
diff --git a/hw/pci/pci.c b/hw/pci/pci.c
index 10b875fdfad1..31ab80b81b6b 100644
--- a/hw/pci/pci.c
+++ b/hw/pci/pci.c
@@ -82,6 +82,8 @@ static Property pci_props[] = {
                     QEMU_PCIE_LNKSTA_DLLLA_BITNR, true),
     DEFINE_PROP_BIT("x-pcie-extcap-init", PCIDevice, cap_present,
                     QEMU_PCIE_EXTCAP_INIT_BITNR, true),
+    DEFINE_PROP_BOOL("unplug-on-migration", PCIDevice,
+                     unplug_on_migration, false),
     DEFINE_PROP_STRING("failover_pair_id", PCIDevice,
                        failover_pair_id),
     DEFINE_PROP_UINT32("acpi-index",  PCIDevice, acpi_index, 0),
@@ -110,6 +112,45 @@ static bool bus_unplug_pending(void *opaque)
     return false;
 }
 
+static int pci_dev_replug_on_migration(void *opaque, QemuOpts *opts,
+                                       Error **errp)
+{
+    Error *err = NULL;
+    const char *bus_name = opaque;
+    const char *opt;
+    DeviceState *dev;
+
+    if (g_strcmp0(qemu_opt_get(opts, "bus"), bus_name)) {
+        return 0;
+    }
+
+    opt = qemu_opt_get(opts, "unplug-on-migration");
+    if (g_strcmp0(opt, "on") && g_strcmp0(opt, "true")) {
+        return 0;
+    }
+    dev = qdev_device_add(opts, &err);
+    if (err) {
+        error_propagate(errp, err);
+        return 1;
+    }
+    object_unref(OBJECT(dev));
+    return 0;
+}
+
+static int bus_post_load(void *opaque, int version_id)
+{
+    Error *err = NULL;
+    PCIBus *bus = opaque;
+
+    if (qemu_opts_foreach(qemu_find_opts("device"),
+                          pci_dev_replug_on_migration, bus->qbus.name, &err)) {
+        error_report_err(err);
+        return -EINVAL;
+    }
+
+    return 0;
+}
+
 static const VMStateDescription vmstate_pcibus = {
     .name = "PCIBUS",
     .version_id = 1,
@@ -122,6 +163,7 @@ static const VMStateDescription vmstate_pcibus = {
         VMSTATE_END_OF_LIST()
     },
     .dev_unplug_pending = bus_unplug_pending,
+    .post_load = bus_post_load,
 };
 
 static void pci_init_bus_master(PCIDevice *pci_dev)
@@ -1200,7 +1242,7 @@ static void pci_qdev_unrealize(DeviceState *dev)
     PCIDevice *pci_dev = PCI_DEVICE(dev);
     PCIDeviceClass *pc = PCI_DEVICE_GET_CLASS(pci_dev);
 
-    if (pci_dev->failover_pair_id) {
+    if (pci_dev->unplug_on_migration) {
         remove_migration_state_change_notifier(&pci_dev->migration_state);
     }
 
@@ -2265,6 +2307,15 @@ static bool pci_dev_hide_device(DeviceListener *listener,
         return false;
     }
 
+    opt = qemu_opt_get(device_opts, "unplug-on-migration");
+    if (g_strcmp0(opt, "on") == 0 || g_strcmp0(opt, "true") == 0) {
+        if (runstate_check(RUN_STATE_INMIGRATE)) {
+            return migration_incoming_get_current()->state !=
+                   MIGRATION_STATUS_ACTIVE;
+        }
+        return false;
+    }
+
     return false;
 }
 
@@ -2290,6 +2341,10 @@ static void pci_qdev_realize(DeviceState *qdev, Error 
**errp)
         pci_dev->cap_present |= QEMU_PCI_CAP_EXPRESS;
     }
 
+    if (pci_dev->failover_pair_id) {
+        pci_dev->unplug_on_migration = true;
+    }
+
     pci_dev = do_pci_register_device(pci_dev,
                                      object_get_typename(OBJECT(qdev)),
                                      pci_dev->devfn, errp);
@@ -2306,12 +2361,6 @@ static void pci_qdev_realize(DeviceState *qdev, Error 
**errp)
     }
 
     if (pci_dev->failover_pair_id) {
-        if (!pci_bus_is_express(pci_get_bus(pci_dev))) {
-            error_setg(errp, "failover primary device must be on "
-                             "PCIExpress bus");
-            pci_qdev_unrealize(DEVICE(pci_dev));
-            return;
-        }
         class_id = pci_get_word(pci_dev->config + PCI_CLASS_DEVICE);
         if (class_id != PCI_CLASS_NETWORK_ETHERNET) {
             error_setg(errp, "failover primary device is not an "
@@ -2319,10 +2368,19 @@ static void pci_qdev_realize(DeviceState *qdev, Error 
**errp)
             pci_qdev_unrealize(DEVICE(pci_dev));
             return;
         }
+    }
+
+    if (pci_dev->unplug_on_migration) {
+        if (!pci_bus_is_express(pci_get_bus(pci_dev))) {
+            error_setg(errp, "Unplugged device on migration must be on "
+                             "PCIExpress bus");
+            pci_qdev_unrealize(DEVICE(pci_dev));
+            return;
+        }
         if ((pci_dev->cap_present & QEMU_PCI_CAP_MULTIFUNCTION)
             || (PCI_FUNC(pci_dev->devfn) != 0)) {
-            error_setg(errp, "failover: primary device must be in its own "
-                              "PCI slot");
+            error_setg(errp, "Unplugged device on migration must be in its "
+                              "own PCI slot");
             pci_qdev_unrealize(DEVICE(pci_dev));
             return;
         }
diff --git a/hw/vfio/pci.c b/hw/vfio/pci.c
index e1ea1d8a23b5..187e1b58a4d9 100644
--- a/hw/vfio/pci.c
+++ b/hw/vfio/pci.c
@@ -3080,7 +3080,7 @@ static void vfio_realize(PCIDevice *pdev, Error **errp)
         }
     }
 
-    if (!pdev->failover_pair_id) {
+    if (!pdev->unplug_on_migration) {
         ret = vfio_migration_probe(&vdev->vbasedev, errp);
         if (ret) {
             error_report("%s: Migration disabled", vdev->vbasedev.name);
-- 
2.31.1