workarounds: Add libvirt_disable_apic

This change adds a new workaround config option to allow disabling the
guest apic.

NOTE(lyarwood): Conflict caused by I66a8e94f6d04cc89785646aa9c3e138f22a51f60
not being present in stable/wallaby (test_driver.py), and
Ie904d1513b5cf76d6d5f6877545e8eb378dd5499 merged in different order in
different branches (workarounds.py).

Conflicts:
    nova/tests/unit/virt/libvirt/test_driver.py
    nova/conf/workarounds.py

Co-authored-by: Lee Yarwood <[email protected]>
Related-Bug: #1939108
Change-Id: I2da867f2734b590a884b1fe1200c402cbf7e9e1c
(cherry picked from commit ec48e15)

Author: 2 people

nova/conf/workarounds.py

@@ -353,6 +353,16 @@
 * :oslo.config:option:`DEFAULT.instances_path`
 * :oslo.config:option:`image_cache.subdirectory_name`
 * :oslo.config:option:`update_resources_interval`
+"""),
+    cfg.BoolOpt('libvirt_disable_apic',
+               default=False,
+               help="""
+With some kernels initializing the guest apic can result in a kernel hang that
+renders the guest unusable. This happens as a result of a kernel bug. In most
+cases the correct fix it to update the guest image kernel to one that is
+patched however in some cases this is not possible. This workaround allows the
+emulation of an apic to be disabled per host however it is not recommended to
+use outside of a CI or developer cloud.
 """),
     cfg.ListOpt('wait_for_vif_plugged_event_during_hard_reboot',
                 item_type=cfg.types.String(

nova/tests/unit/virt/libvirt/test_driver.py

@@ -4909,6 +4909,33 @@ def _test_get_guest_config_windows_hyperv(
         self.assertTrue(cfg.features[2].vapic)
         self.assertEqual(hvid_hidden, cfg.features[2].vendorid_spoof)
 
+    @mock.patch.object(host.Host, 'has_min_version',
+                       new=mock.Mock(return_value=True))
+    def test_get_guest_config_apic_workaround(self):
+        self.flags(virt_type='qemu', group='libvirt')
+        drvr = libvirt_driver.LibvirtDriver(fake.FakeVirtAPI(), True)
+        instance_ref = objects.Instance(**self.test_instance)
+        image_meta = objects.ImageMeta.from_dict(self.test_image_meta)
+        disk_info = blockinfo.get_disk_info(
+            CONF.libvirt.virt_type, instance_ref, image_meta)
+
+        cfg = drvr._get_guest_config(
+            instance_ref, _fake_network_info(self), image_meta, disk_info)
+
+        self.assertEqual(2, len(cfg.features))
+        self.assertIsInstance(
+            cfg.features[0], vconfig.LibvirtConfigGuestFeatureACPI)
+        self.assertIsInstance(
+            cfg.features[1], vconfig.LibvirtConfigGuestFeatureAPIC)
+
+        self.flags(libvirt_disable_apic=True, group='workarounds')
+        cfg = drvr._get_guest_config(
+            instance_ref, _fake_network_info(self), image_meta, disk_info)
+
+        self.assertEqual(1, len(cfg.features))
+        self.assertIsInstance(
+            cfg.features[0], vconfig.LibvirtConfigGuestFeatureACPI)
+
     def test_get_guest_config_windows_hyperv_feature2(self):
         self._test_get_guest_config_windows_hyperv()
 

nova/virt/libvirt/driver.py

@@ -5867,7 +5867,8 @@ def _set_features(self, guest, os_type, image_meta, flavor):
 
         if CONF.libvirt.virt_type in ('qemu', 'kvm'):
             guest.features.append(vconfig.LibvirtConfigGuestFeatureACPI())
-            guest.features.append(vconfig.LibvirtConfigGuestFeatureAPIC())
+            if not CONF.workarounds.libvirt_disable_apic:
+                guest.features.append(vconfig.LibvirtConfigGuestFeatureAPIC())
 
         if CONF.libvirt.virt_type in ('qemu', 'kvm') and os_type == 'windows':
             hv = vconfig.LibvirtConfigGuestFeatureHyperV()

releasenotes/notes/libvirt-disable-apic-39599bdc2d110a1f.yaml

@@ -0,0 +1,13 @@
+---
+issues:
+  - |
+    Linux guest images that have known kernel bugs related to virtualized apic
+    initialization previously would sporadically hang. For images where the
+    kernel cannot be upgraded, a ``[workarounds]`` config option has been
+    introduced:
+
+    ``[workarounds]libvirt_disable_apic``
+
+    This option is primarily intended for CI and development clouds as a bridge
+    for operators to mitigate the issue while they work with their upstream
+    image vendors.