[PATCH v2 16/16] qemu: leave restricting cpuset.mems after initialization

[Martin Kletzander <mkletzan@xxxxxxxxxx>]

When domain is started with numatune memory mode strict and the
nodeset does not include host NUMA node with DMA and DMA32 zones, KVM
initialization fails.  This is because cgroup restrict even kernel
allocations.  We are already doing numa_set_membind() which does the
same thing, only it does not restrict kernel allocations.

This patch leaves the userspace numa_set_membind() in place and moves
the cpuset.mems setting after the point where monitor comes up, but
before vcpu and emulator sub-groups are created.

Signed-off-by: Martin Kletzander <mkletzan@xxxxxxxxxx>
---

Notes:
    Another approach would be not using cgroups for this at all; it should
    still work as expected.

 src/qemu/qemu_cgroup.c  | 10 +++++++---
 src/qemu/qemu_cgroup.h  |  4 +++-
 src/qemu/qemu_process.c |  4 ++++
 3 files changed, 14 insertions(+), 4 deletions(-)

diff --git a/src/qemu/qemu_cgroup.c b/src/qemu/qemu_cgroup.c
index eebe9e9..ffef1fb 100644
--- a/src/qemu/qemu_cgroup.c
+++ b/src/qemu/qemu_cgroup.c
@@ -614,9 +614,6 @@ qemuSetupCpusetCgroup(virDomainObjPtr vm,
     if (!virCgroupHasController(priv->cgroup, VIR_CGROUP_CONTROLLER_CPUSET))
         return 0;

-    if (qemuSetupCpusetMems(vm, nodemask) < 0)
-        goto cleanup;
-
     if (vm->def->cpumask ||
         (vm->def->placement_mode == VIR_DOMAIN_CPU_PLACEMENT_MODE_AUTO)) {

@@ -813,6 +810,13 @@ qemuSetupCgroup(virQEMUDriverPtr driver,
 }

 int
+qemuSetupCgroupPostInit(virDomainObjPtr vm,
+                        virBitmapPtr nodemask)
+{
+    return qemuSetupCpusetMems(vm, nodemask);
+}
+
+int
 qemuSetupCgroupVcpuBW(virCgroupPtr cgroup,
                       unsigned long long period,
                       long long quota)
diff --git a/src/qemu/qemu_cgroup.h b/src/qemu/qemu_cgroup.h
index 14404d1..40a031e 100644
--- a/src/qemu/qemu_cgroup.h
+++ b/src/qemu/qemu_cgroup.h
@@ -1,7 +1,7 @@
 /*
  * qemu_cgroup.h: QEMU cgroup management
  *
- * Copyright (C) 2006-2007, 2009-2013 Red Hat, Inc.
+ * Copyright (C) 2006-2007, 2009-2014 Red Hat, Inc.
  * Copyright (C) 2006 Daniel P. Berrange
  *
  * This library is free software; you can redistribute it and/or
@@ -44,6 +44,8 @@ int qemuConnectCgroup(virQEMUDriverPtr driver,
 int qemuSetupCgroup(virQEMUDriverPtr driver,
                     virDomainObjPtr vm,
                     virBitmapPtr nodemask);
+int qemuSetupCgroupPostInit(virDomainObjPtr vm,
+                            virBitmapPtr nodemask);
 int qemuSetupCgroupVcpuBW(virCgroupPtr cgroup,
                           unsigned long long period,
                           long long quota);
diff --git a/src/qemu/qemu_process.c b/src/qemu/qemu_process.c
index 4a27eab..8e70258 100644
--- a/src/qemu/qemu_process.c
+++ b/src/qemu/qemu_process.c
@@ -4166,6 +4166,10 @@ int qemuProcessStart(virConnectPtr conn,
     if (!qemuProcessVerifyGuestCPU(driver, vm))
         goto cleanup;

+    VIR_DEBUG("Setting up post-init cgroup restrictions");
+    if (qemuSetupCgroupPostInit(vm, nodemask) < 0)
+        goto cleanup;
+
     VIR_DEBUG("Detecting VCPU PIDs");
     if (qemuProcessDetectVcpuPIDs(driver, vm) < 0)
         goto cleanup;
-- 
2.0.0

--
libvir-list mailing list
libvir-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/libvir-list