qemu: Avoid dangling migration-out job when client dies

When a client which started non-p2p migration dies in a bad time, the
source libvirtd never clears the migration job and almost nothing can be
done with the domain without restarting the daemon. This patch makes use
of connection close callbacks and ensures that migration job is properly
discarded when the client disconnects.

diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c
index 4c272251cfbfde29ac1a08aa0341ef4b251ea5e5..b00257ba8aaf5dbaa7297cae63078b593389c1f5 100644 (file)
--- a/src/qemu/qemu_driver.c
+++ b/src/qemu/qemu_driver.c
@@ -8838,6 +8838,9 @@ qemuDomainMigrateBegin3(virDomainPtr domain,
          * This prevents any other APIs being invoked while migration is taking
          * place.
          */
+        if (qemuDriverCloseCallbackSet(driver, vm, domain->conn,
+                                       qemuMigrationCleanup) < 0)
+            goto endjob;
         if (qemuMigrationJobContinue(vm) == 0) {
             vm = NULL;
             qemuReportError(VIR_ERR_OPERATION_FAILED,
@@ -9069,6 +9072,7 @@ qemuDomainMigrateConfirm3(virDomainPtr domain,
         phase = QEMU_MIGRATION_PHASE_CONFIRM3;
 
     qemuMigrationJobStartPhase(driver, vm, phase);
+    qemuDriverCloseCallbackUnset(driver, vm, qemuMigrationCleanup);
 
     ret = qemuMigrationConfirm(driver, domain->conn, vm,
                                cookiein, cookieinlen,

diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c
index 69d4f0c52066357d9f6fb83c3f83b8106bf4e22a..12ce10c86fc1aaa3bb7fce717c3ad978912bdb76 100644 (file)
--- a/src/qemu/qemu_migration.c
+++ b/src/qemu/qemu_migration.c
@@ -1037,6 +1037,67 @@ qemuDomainMigrateGraphicsRelocate(struct qemud_driver *driver,
 }
 
 
+/* This is called for outgoing non-p2p migrations when a connection to the
+ * client which initiated the migration was closed but we were waiting for it
+ * to follow up with the next phase, that is, in between
+ * qemuDomainMigrateBegin3 and qemuDomainMigratePerform3 or
+ * qemuDomainMigratePerform3 and qemuDomainMigrateConfirm3.
+ */
+virDomainObjPtr
+qemuMigrationCleanup(struct qemud_driver *driver,
+                     virDomainObjPtr vm,
+                     virConnectPtr conn)
+{
+    qemuDomainObjPrivatePtr priv = vm->privateData;
+
+    VIR_DEBUG("vm=%s, conn=%p, asyncJob=%s, phase=%s",
+              vm->def->name, conn,
+              qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
+              qemuDomainAsyncJobPhaseToString(priv->job.asyncJob,
+                                              priv->job.phase));
+
+    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_OUT))
+        goto cleanup;
+
+    VIR_DEBUG("The connection which started outgoing migration of domain %s"
+              " was closed; canceling the migration",
+              vm->def->name);
+
+    switch ((enum qemuMigrationJobPhase) priv->job.phase) {
+    case QEMU_MIGRATION_PHASE_BEGIN3:
+        /* just forget we were about to migrate */
+        qemuDomainObjDiscardAsyncJob(driver, vm);
+        break;
+
+    case QEMU_MIGRATION_PHASE_PERFORM3_DONE:
+        VIR_WARN("Migration of domain %s finished but we don't know if the"
+                 " domain was successfully started on destination or not",
+                 vm->def->name);
+        /* clear the job and let higher levels decide what to do */
+        qemuDomainObjDiscardAsyncJob(driver, vm);
+        break;
+
+    case QEMU_MIGRATION_PHASE_PERFORM3:
+        /* cannot be seen without an active migration API; unreachable */
+    case QEMU_MIGRATION_PHASE_CONFIRM3:
+    case QEMU_MIGRATION_PHASE_CONFIRM3_CANCELLED:
+        /* all done; unreachable */
+    case QEMU_MIGRATION_PHASE_PREPARE:
+    case QEMU_MIGRATION_PHASE_FINISH2:
+    case QEMU_MIGRATION_PHASE_FINISH3:
+        /* incoming migration; unreachable */
+    case QEMU_MIGRATION_PHASE_PERFORM2:
+        /* single phase outgoing migration; unreachable */
+    case QEMU_MIGRATION_PHASE_NONE:
+    case QEMU_MIGRATION_PHASE_LAST:
+        /* unreachable */
+        ;
+    }
+
+cleanup:
+    return vm;
+}
+
 /* The caller is supposed to lock the vm and start a migration job. */
 char *qemuMigrationBegin(struct qemud_driver *driver,
                          virDomainObjPtr vm,
@@ -2547,6 +2608,7 @@ qemuMigrationPerformPhase(struct qemud_driver *driver,
     }
 
     qemuMigrationJobStartPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
+    qemuDriverCloseCallbackUnset(driver, vm, qemuMigrationCleanup);
 
     resume = virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING;
     ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
@@ -2577,6 +2639,10 @@ qemuMigrationPerformPhase(struct qemud_driver *driver,
 
     qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3_DONE);
 
+    if (qemuDriverCloseCallbackSet(driver, vm, conn,
+                                   qemuMigrationCleanup) < 0)
+        goto endjob;
+
 endjob:
     if (ret < 0)
         refs = qemuMigrationJobFinish(driver, vm);

diff --git a/src/qemu/qemu_migration.h b/src/qemu/qemu_migration.h
index 41e4eac9dddf9191db8aafa6cf3b96958ef10446..5fab0ca88fb40182659b251681435e6b1e947627 100644 (file)
--- a/src/qemu/qemu_migration.h
+++ b/src/qemu/qemu_migration.h
@@ -77,6 +77,10 @@ int qemuMigrationJobFinish(struct qemud_driver *driver, virDomainObjPtr obj)
 int qemuMigrationSetOffline(struct qemud_driver *driver,
                             virDomainObjPtr vm);
 
+virDomainObjPtr qemuMigrationCleanup(struct qemud_driver *driver,
+                                     virDomainObjPtr vm,
+                                     virConnectPtr conn);
+
 char *qemuMigrationBegin(struct qemud_driver *driver,
                          virDomainObjPtr vm,
                          const char *xmlin,