Change default cgroup layout for QEMU/LXC and honour XML config

Historically QEMU/LXC guests have been placed in a cgroup layout
that is

   $LOCATION-OF-LIBVIRTD/libvirt/{qemu,lxc}/$VMNAME

This is bad for a number of reasons

 - The cgroup hierarchy gets very deep which seriously
   impacts kernel performance due to cgroups scalability
   limitations.

 - It is hard to setup cgroup policies which apply across
   services and virtual machines, since all VMs are underneath
   the libvirtd service.

To address this the default cgroup location is changed to
be

    /system/$VMNAME.{lxc,qemu}.libvirt

This puts virtual machines at the same level in the hierarchy
as system services, allowing consistent policy to be setup
across all of them.

This also honours the new resource partition location from the
XML configuration, for example

  <resource>
    <partition>/virtualmachines/production</partitions>
  </resource>

will result in the VM being placed at

    /virtualmachines/production/$VMNAME.{lxc,qemu}.libvirt

NB, with the exception of the default, /system, path which
is intended to always exist, libvirt will not attempt to
auto-create the partitions in the XML. It is the responsibility
of the admin/app to configure the partitions. Later libvirt
APIs will provide a way todo this.

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>

src/lxc/lxc_cgroup.c

@@ -523,29 +523,88 @@ cleanup:
 }
 
 
-virCgroupPtr virLXCCgroupCreate(virDomainDefPtr def)
+virCgroupPtr virLXCCgroupCreate(virDomainDefPtr def, bool startup)
 {
-    virCgroupPtr driver = NULL;
-    virCgroupPtr cgroup = NULL;
     int rc;
+    virCgroupPtr parent = NULL;
+    virCgroupPtr cgroup = NULL;
 
-    rc = virCgroupNewDriver("lxc", true, false, -1, &driver);
+    if (!def->resource && startup) {
-    if (rc != 0) {
+        virDomainResourceDefPtr res;
-        virReportSystemError(-rc, "%s",
+
-                             _("Unable to get cgroup for driver"));
+        if (VIR_ALLOC(res) < 0) {
+            virReportOOMError();
             goto cleanup;
         }
 
-    rc = virCgroupNewDomainDriver(driver, def->name, true, &cgroup);
+        if (!(res->partition = strdup("/system"))) {
+            virReportOOMError();
+            VIR_FREE(res);
+            goto cleanup;
+        }
+
+        def->resource = res;
+    }
+
+    if (def->resource &&
+        def->resource->partition) {
+        if (def->resource->partition[0] != '/') {
+            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
+                           _("Resource partition '%s' must start with '/'"),
+                           def->resource->partition);
+            goto cleanup;
+        }
+        /* We only auto-create the default partition. In other
+         * cases we expec the sysadmin/app to have done so */
+        rc = virCgroupNewPartition(def->resource->partition,
+                                   STREQ(def->resource->partition, "/system"),
+                                   -1,
+                                   &parent);
         if (rc != 0) {
             virReportSystemError(-rc,
-                             _("Unable to create cgroup for domain %s"),
+                                 _("Unable to initialize %s cgroup"),
+                                 def->resource->partition);
+            goto cleanup;
+        }
+
+        rc = virCgroupNewDomainPartition(parent,
+                                         "lxc",
+                                         def->name,
+                                         true,
+                                         &cgroup);
+        if (rc != 0) {
+            virReportSystemError(-rc,
+                                 _("Unable to create cgroup for %s"),
+                                 def->name);
+            goto cleanup;
+        }
+    } else {
+        rc = virCgroupNewDriver("lxc",
+                                true,
+                                true,
+                                -1,
+                                &parent);
+        if (rc != 0) {
+            virReportSystemError(-rc,
+                                 _("Unable to create cgroup for %s"),
                                  def->name);
             goto cleanup;
         }
 
+        rc = virCgroupNewDomainDriver(parent,
+                                      def->name,
+                                      true,
+                                      &cgroup);
+        if (rc != 0) {
+            virReportSystemError(-rc,
+                                 _("Unable to create cgroup for %s"),
+                                 def->name);
+            goto cleanup;
+        }
+    }
+
 cleanup:
-    virCgroupFree(&driver);
+    virCgroupFree(&parent);
     return cgroup;
 }
 
@@ -556,7 +615,7 @@ virCgroupPtr virLXCCgroupJoin(virDomainDefPtr def)
     int ret = -1;
     int rc;
 
-    if (!(cgroup = virLXCCgroupCreate(def)))
+    if (!(cgroup = virLXCCgroupCreate(def, true)))
         return NULL;
 
     rc = virCgroupAddTask(cgroup, getpid());

src/lxc/lxc_cgroup.h

@@ -27,7 +27,7 @@
 # include "lxc_fuse.h"
 # include "virusb.h"
 
-virCgroupPtr virLXCCgroupCreate(virDomainDefPtr def);
+virCgroupPtr virLXCCgroupCreate(virDomainDefPtr def, bool startup);
 virCgroupPtr virLXCCgroupJoin(virDomainDefPtr def);
 int virLXCCgroupSetup(virDomainDefPtr def,
                       virCgroupPtr cgroup,

src/lxc/lxc_process.c

@@ -1049,7 +1049,7 @@ int virLXCProcessStart(virConnectPtr conn,
 
     virCgroupFree(&priv->cgroup);
 
-    if (!(priv->cgroup = virLXCCgroupCreate(vm->def)))
+    if (!(priv->cgroup = virLXCCgroupCreate(vm->def, true)))
         return -1;
 
     if (!virCgroupHasController(priv->cgroup,
@@ -1464,7 +1464,7 @@ virLXCProcessReconnectDomain(virDomainObjPtr vm,
         if (!(priv->monitor = virLXCProcessConnectMonitor(driver, vm)))
             goto error;
 
-        if (!(priv->cgroup = virLXCCgroupCreate(vm->def)))
+        if (!(priv->cgroup = virLXCCgroupCreate(vm->def, false)))
             goto error;
 
         if (virLXCUpdateActiveUsbHostdevs(driver, vm->def) < 0)

src/qemu/qemu_cgroup.c

@@ -215,20 +215,78 @@ int qemuSetupHostUsbDeviceCgroup(virUSBDevicePtr dev ATTRIBUTE_UNUSED,
 
 
 int qemuInitCgroup(virQEMUDriverPtr driver,
-                   virDomainObjPtr vm)
+                   virDomainObjPtr vm,
+                   bool startup)
 {
-    int rc;
+    int rc = -1;
     qemuDomainObjPrivatePtr priv = vm->privateData;
-    virCgroupPtr driverGroup = NULL;
+    virCgroupPtr parent = NULL;
     virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
 
     virCgroupFree(&priv->cgroup);
 
+    if (!vm->def->resource && startup) {
+        virDomainResourceDefPtr res;
+
+        if (VIR_ALLOC(res) < 0) {
+            virReportOOMError();
+            goto cleanup;
+        }
+
+        if (!(res->partition = strdup("/system"))) {
+            virReportOOMError();
+            VIR_FREE(res);
+            goto cleanup;
+        }
+
+        vm->def->resource = res;
+    }
+
+    if (vm->def->resource &&
+        vm->def->resource->partition) {
+        if (vm->def->resource->partition[0] != '/') {
+            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
+                           _("Resource partition '%s' must start with '/'"),
+                           vm->def->resource->partition);
+            goto cleanup;
+        }
+        /* We only auto-create the default partition. In other
+         * cases we expec the sysadmin/app to have done so */
+        rc = virCgroupNewPartition(vm->def->resource->partition,
+                                   STREQ(vm->def->resource->partition, "/system"),
+                                   cfg->cgroupControllers,
+                                   &parent);
+        if (rc != 0) {
+            if (rc == -ENXIO ||
+                rc == -EPERM ||
+                rc == -EACCES) { /* No cgroups mounts == success */
+                VIR_DEBUG("No cgroups present/configured/accessible, ignoring error");
+                goto done;
+            }
+
+            virReportSystemError(-rc,
+                                 _("Unable to initialize %s cgroup"),
+                                 vm->def->resource->partition);
+            goto cleanup;
+        }
+
+        rc = virCgroupNewDomainPartition(parent,
+                                         "qemu",
+                                         vm->def->name,
+                                         true,
+                                         &priv->cgroup);
+        if (rc != 0) {
+            virReportSystemError(-rc,
+                                 _("Unable to create cgroup for %s"),
+                                 vm->def->name);
+            goto cleanup;
+        }
+    } else {
         rc = virCgroupNewDriver("qemu",
                                 cfg->privileged,
                                 true,
                                 cfg->cgroupControllers,
-                            &driverGroup);
+                                &parent);
         if (rc != 0) {
             if (rc == -ENXIO ||
                 rc == -EPERM ||
@@ -243,18 +301,22 @@ int qemuInitCgroup(virQEMUDriverPtr driver,
             goto cleanup;
         }
 
-    rc = virCgroupNewDomainDriver(driverGroup, vm->def->name, true, &priv->cgroup);
+        rc = virCgroupNewDomainDriver(parent,
+                                      vm->def->name,
+                                      true,
+                                      &priv->cgroup);
         if (rc != 0) {
             virReportSystemError(-rc,
                                  _("Unable to create cgroup for %s"),
                                  vm->def->name);
             goto cleanup;
         }
+    }
 
 done:
     rc = 0;
 cleanup:
-    virCgroupFree(&driverGroup);
+    virCgroupFree(&parent);
     virObjectUnref(cfg);
     return rc;
 }
@@ -273,7 +335,7 @@ int qemuSetupCgroup(virQEMUDriverPtr driver,
         (const char *const *)cfg->cgroupDeviceACL :
         defaultDeviceACL;
 
-    if (qemuInitCgroup(driver, vm) < 0)
+    if (qemuInitCgroup(driver, vm, true) < 0)
         return -1;
 
     if (!priv->cgroup)

src/qemu/qemu_cgroup.h

@@ -37,7 +37,8 @@ int qemuSetupHostUsbDeviceCgroup(virUSBDevicePtr dev,
                                  const char *path,
                                  void *opaque);
 int qemuInitCgroup(virQEMUDriverPtr driver,
-                   virDomainObjPtr vm);
+                   virDomainObjPtr vm,
+                   bool startup);
 int qemuSetupCgroup(virQEMUDriverPtr driver,
                     virDomainObjPtr vm,
                     virBitmapPtr nodemask);

src/qemu/qemu_process.c

@@ -3005,7 +3005,7 @@ qemuProcessReconnect(void *opaque)
     if (qemuUpdateActiveUsbHostdevs(driver, obj->def) < 0)
         goto error;
 
-    if (qemuInitCgroup(driver, obj) < 0)
+    if (qemuInitCgroup(driver, obj, false) < 0)
         goto error;
 
     /* XXX: Need to change as long as lock is introduced for