int
qemuSetupCpusetMems(virDomainObjPtr vm)
{
virCgroupPtr cgroup_temp = NULL;
qemuDomainObjPrivatePtr priv = vm->privateData;
virDomainNumatuneMemMode mode;
char *mem_mask = NULL;
int ret = -1;
if (!virCgroupHasController(priv->cgroup, VIR_CGROUP_CONTROLLER_CPUSET))
return 0;
if (virDomainNumatuneGetMode(vm->def->numa, -1, &mode) < 0 ||
mode != VIR_DOMAIN_NUMATUNE_MEM_STRICT)
return 0;
if (virDomainNumatuneMaybeFormatNodeset(vm->def->numa,
priv->autoNodeset,
&mem_mask, -1) < 0)
goto cleanup;
if (mem_mask)
if (virCgroupNewThread(priv->cgroup, VIR_CGROUP_THREAD_EMULATOR, 0,
false, &cgroup_temp) < 0 ||
virCgroupSetCpusetMems(cgroup_temp, mem_mask) < 0)
goto cleanup;
ret = 0;
cleanup:
VIR_FREE(mem_mask);
virCgroupFree(&cgroup_temp);
return ret;
}
static int virLXCCgroupSetupCpusetTune(virDomainDefPtr def,
virCgroupPtr cgroup,
virBitmapPtr nodemask)
{
int ret = -1;
char *mask = NULL;
if (def->placement_mode != VIR_DOMAIN_CPU_PLACEMENT_MODE_AUTO &&
def->cpumask) {
if (!(mask = virBitmapFormat(def->cpumask)))
return -1;
if (virCgroupSetCpusetCpus(cgroup, mask) < 0)
goto cleanup;
}
if (virDomainNumatuneGetMode(def->numa, -1) !=
VIR_DOMAIN_NUMATUNE_MEM_STRICT)
goto cleanup;
if (virDomainNumatuneMaybeFormatNodeset(def->numa, nodemask,
&mask, -1) < 0)
goto cleanup;
if (mask && virCgroupSetCpusetMems(cgroup, mask) < 0)
goto cleanup;
ret = 0;
cleanup:
VIR_FREE(mask);
return ret;
}
int
qemuSetupGlobalCpuCgroup(virDomainObjPtr vm)
{
qemuDomainObjPrivatePtr priv = vm->privateData;
unsigned long long period = vm->def->cputune.global_period;
long long quota = vm->def->cputune.global_quota;
char *mem_mask = NULL;
virDomainNumatuneMemMode mem_mode;
if ((period || quota) &&
!virCgroupHasController(priv->cgroup, VIR_CGROUP_CONTROLLER_CPU)) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
_("cgroup cpu is required for scheduler tuning"));
return -1;
}
/*
* If CPU cgroup controller is not initialized here, then we need
* neither period nor quota settings. And if CPUSET controller is
* not initialized either, then there's nothing to do anyway.
*/
if (!virCgroupHasController(priv->cgroup, VIR_CGROUP_CONTROLLER_CPU) &&
!virCgroupHasController(priv->cgroup, VIR_CGROUP_CONTROLLER_CPUSET))
return 0;
if (virDomainNumatuneGetMode(vm->def->numa, -1, &mem_mode) == 0 &&
mem_mode == VIR_DOMAIN_NUMATUNE_MEM_STRICT &&
virDomainNumatuneMaybeFormatNodeset(vm->def->numa,
priv->autoNodeset,
&mem_mask, -1) < 0)
goto cleanup;
if (period || quota) {
if (qemuSetupCgroupVcpuBW(priv->cgroup, period, quota) < 0)
goto cleanup;
}
VIR_FREE(mem_mask);
return 0;
cleanup:
VIR_FREE(mem_mask);
return -1;
}
int
virNumaSetupMemoryPolicy(virDomainNumatunePtr numatune,
virBitmapPtr nodemask)
{
nodemask_t mask;
int node = -1;
int ret = -1;
int bit = 0;
size_t i;
int maxnode = 0;
virBitmapPtr tmp_nodemask = NULL;
tmp_nodemask = virDomainNumatuneGetNodeset(numatune, nodemask, -1);
if (!tmp_nodemask)
return 0;
if (numa_available() < 0) {
virReportError(VIR_ERR_INTERNAL_ERROR,
"%s", _("Host kernel is not aware of NUMA."));
return -1;
}
maxnode = numa_max_node();
maxnode = maxnode < NUMA_NUM_NODES ? maxnode : NUMA_NUM_NODES;
/* Convert nodemask to NUMA bitmask. */
nodemask_zero(&mask);
bit = -1;
while ((bit = virBitmapNextSetBit(tmp_nodemask, bit)) >= 0) {
if (bit > maxnode) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("NUMA node %d is out of range"), bit);
return -1;
}
nodemask_set(&mask, bit);
}
switch (virDomainNumatuneGetMode(numatune, -1)) {
case VIR_DOMAIN_NUMATUNE_MEM_STRICT:
numa_set_bind_policy(1);
numa_set_membind(&mask);
numa_set_bind_policy(0);
break;
case VIR_DOMAIN_NUMATUNE_MEM_PREFERRED:
{
int nnodes = 0;
for (i = 0; i < NUMA_NUM_NODES; i++) {
if (nodemask_isset(&mask, i)) {
node = i;
nnodes++;
}
}
if (nnodes != 1) {
virReportError(VIR_ERR_INTERNAL_ERROR,
"%s", _("NUMA memory tuning in 'preferred' mode "
"only supports single node"));
goto cleanup;
}
numa_set_bind_policy(0);
numa_set_preferred(node);
}
break;
case VIR_DOMAIN_NUMATUNE_MEM_INTERLEAVE:
numa_set_interleave_mask(&mask);
break;
case VIR_DOMAIN_NUMATUNE_MEM_LAST:
break;
}
ret = 0;
cleanup:
return ret;
}
int
qemuSetupCgroupForVcpu(virDomainObjPtr vm)
{
virCgroupPtr cgroup_vcpu = NULL;
qemuDomainObjPrivatePtr priv = vm->privateData;
virDomainDefPtr def = vm->def;
size_t i, j;
unsigned long long period = vm->def->cputune.period;
long long quota = vm->def->cputune.quota;
char *mem_mask = NULL;
virDomainNumatuneMemMode mem_mode;
if ((period || quota) &&
!virCgroupHasController(priv->cgroup, VIR_CGROUP_CONTROLLER_CPU)) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
_("cgroup cpu is required for scheduler tuning"));
return -1;
}
/*
* If CPU cgroup controller is not initialized here, then we need
* neither period nor quota settings. And if CPUSET controller is
* not initialized either, then there's nothing to do anyway.
*/
if (!virCgroupHasController(priv->cgroup, VIR_CGROUP_CONTROLLER_CPU) &&
!virCgroupHasController(priv->cgroup, VIR_CGROUP_CONTROLLER_CPUSET))
return 0;
/* We are trying to setup cgroups for CPU pinning, which can also be done
* with virProcessSetAffinity, thus the lack of cgroups is not fatal here.
*/
if (priv->cgroup == NULL)
return 0;
if (priv->nvcpupids == 0 || priv->vcpupids[0] == vm->pid) {
/* If we don't know VCPU<->PID mapping or all vcpu runs in the same
* thread, we cannot control each vcpu.
*/
return 0;
}
if (virDomainNumatuneGetMode(vm->def->numa, -1, &mem_mode) == 0 &&
mem_mode == VIR_DOMAIN_NUMATUNE_MEM_STRICT &&
virDomainNumatuneMaybeFormatNodeset(vm->def->numa,
priv->autoNodeset,
&mem_mask, -1) < 0)
goto cleanup;
for (i = 0; i < priv->nvcpupids; i++) {
virCgroupFree(&cgroup_vcpu);
if (virCgroupNewThread(priv->cgroup, VIR_CGROUP_THREAD_VCPU, i,
true, &cgroup_vcpu) < 0)
goto cleanup;
/* move the thread for vcpu to sub dir */
if (virCgroupAddTask(cgroup_vcpu, priv->vcpupids[i]) < 0)
goto cleanup;
if (period || quota) {
if (qemuSetupCgroupVcpuBW(cgroup_vcpu, period, quota) < 0)
goto cleanup;
}
/* Set vcpupin in cgroup if vcpupin xml is provided */
if (virCgroupHasController(priv->cgroup, VIR_CGROUP_CONTROLLER_CPUSET)) {
virBitmapPtr cpumap = NULL;
if (mem_mask &&
virCgroupSetCpusetMems(cgroup_vcpu, mem_mask) < 0)
goto cleanup;
/* try to use the default cpu maps */
if (vm->def->placement_mode == VIR_DOMAIN_CPU_PLACEMENT_MODE_AUTO)
cpumap = priv->autoCpuset;
else
cpumap = vm->def->cpumask;
/* lookup a more specific pinning info */
for (j = 0; j < def->cputune.nvcpupin; j++) {
if (def->cputune.vcpupin[j]->id == i) {
cpumap = def->cputune.vcpupin[j]->cpumask;
break;
}
}
if (!cpumap)
continue;
if (qemuSetupCgroupCpusetCpus(cgroup_vcpu, cpumap) < 0)
goto cleanup;
}
}
virCgroupFree(&cgroup_vcpu);
VIR_FREE(mem_mask);
return 0;
cleanup:
if (cgroup_vcpu) {
virCgroupRemove(cgroup_vcpu);
virCgroupFree(&cgroup_vcpu);
}
VIR_FREE(mem_mask);
return -1;
}
int
qemuSetupCgroupForIOThreads(virDomainObjPtr vm)
{
virCgroupPtr cgroup_iothread = NULL;
qemuDomainObjPrivatePtr priv = vm->privateData;
virDomainDefPtr def = vm->def;
size_t i;
unsigned long long period = vm->def->cputune.period;
long long quota = vm->def->cputune.quota;
char *mem_mask = NULL;
virDomainNumatuneMemMode mem_mode;
if ((period || quota) &&
!virCgroupHasController(priv->cgroup, VIR_CGROUP_CONTROLLER_CPU)) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
_("cgroup cpu is required for scheduler tuning"));
return -1;
}
/*
* If CPU cgroup controller is not initialized here, then we need
* neither period nor quota settings. And if CPUSET controller is
* not initialized either, then there's nothing to do anyway.
*/
if (!virCgroupHasController(priv->cgroup, VIR_CGROUP_CONTROLLER_CPU) &&
!virCgroupHasController(priv->cgroup, VIR_CGROUP_CONTROLLER_CPUSET))
return 0;
/* We are trying to setup cgroups for CPU pinning, which can also be done
* with virProcessSetAffinity, thus the lack of cgroups is not fatal here.
*/
if (priv->cgroup == NULL)
return 0;
if (virDomainNumatuneGetMode(vm->def->numa, -1, &mem_mode) == 0 &&
mem_mode == VIR_DOMAIN_NUMATUNE_MEM_STRICT &&
virDomainNumatuneMaybeFormatNodeset(vm->def->numa,
priv->autoNodeset,
&mem_mask, -1) < 0)
goto cleanup;
for (i = 0; i < def->niothreadids; i++) {
/* IOThreads are numbered 1..n, although the array is 0..n-1,
* so we will account for that here
*/
if (virCgroupNewThread(priv->cgroup, VIR_CGROUP_THREAD_IOTHREAD,
def->iothreadids[i]->iothread_id,
true, &cgroup_iothread) < 0)
goto cleanup;
/* move the thread for iothread to sub dir */
if (virCgroupAddTask(cgroup_iothread,
def->iothreadids[i]->thread_id) < 0)
goto cleanup;
if (period || quota) {
if (qemuSetupCgroupVcpuBW(cgroup_iothread, period, quota) < 0)
goto cleanup;
}
/* Set iothreadpin in cgroup if iothreadpin xml is provided */
if (virCgroupHasController(priv->cgroup,
VIR_CGROUP_CONTROLLER_CPUSET)) {
virBitmapPtr cpumask = NULL;
if (mem_mask &&
virCgroupSetCpusetMems(cgroup_iothread, mem_mask) < 0)
goto cleanup;
if (def->iothreadids[i]->cpumask)
cpumask = def->iothreadids[i]->cpumask;
else if (def->placement_mode == VIR_DOMAIN_CPU_PLACEMENT_MODE_AUTO)
cpumask = priv->autoCpuset;
else
cpumask = def->cpumask;
if (cpumask &&
qemuSetupCgroupCpusetCpus(cgroup_iothread, cpumask) < 0)
goto cleanup;
}
virCgroupFree(&cgroup_iothread);
}
VIR_FREE(mem_mask);
return 0;
cleanup:
if (cgroup_iothread) {
virCgroupRemove(cgroup_iothread);
virCgroupFree(&cgroup_iothread);
}
VIR_FREE(mem_mask);
return -1;
}
