On Thu, Nov 17, 2011 at 05:44:20PM +0800, Hu Tao wrote:
> ---
> src/qemu/qemu_driver.c | 399 ++++++++++++++++++++++++++++++++++++++++++++++++
> 1 files changed, 399 insertions(+), 0 deletions(-)
>
> diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c
> index 5f4a18d..5b6398c 100644
> --- a/src/qemu/qemu_driver.c
> +++ b/src/qemu/qemu_driver.c
> @@ -94,6 +94,8 @@
>
> #define QEMU_NB_MEM_PARAM 3
>
> +#define QEMU_NB_NUMA_PARAM 2
> +
> #if HAVE_LINUX_KVM_H
> # include <linux/kvm.h>
> #endif
> @@ -6524,6 +6526,401 @@ cleanup:
> return ret;
> }
>
> +static int qemuDomainSetNumaParameters(virDomainPtr dom,
> + virTypedParameterPtr params,
> + int nparams,
> + unsigned int flags)
> +{
> + struct qemud_driver *driver = dom->conn->privateData;
> + int i;
> + virDomainDefPtr persistentDef = NULL;
> + virCgroupPtr group = NULL;
> + virDomainObjPtr vm = NULL;
> + int ret = -1;
> + bool isActive;
> +
> + virCheckFlags(VIR_DOMAIN_AFFECT_LIVE |
> + VIR_DOMAIN_AFFECT_CONFIG, -1);
> +
> + qemuDriverLock(driver);
> +
> + vm = virDomainFindByUUID(&driver->domains, dom->uuid);
> +
> + if (vm == NULL) {
> + qemuReportError(VIR_ERR_INTERNAL_ERROR,
> + _("No such domain %s"), dom->uuid);
> + goto cleanup;
> + }
> +
> + isActive = virDomainObjIsActive(vm);
> +
> + if (flags == VIR_DOMAIN_AFFECT_CURRENT) {
> + if (isActive)
> + flags = VIR_DOMAIN_AFFECT_LIVE;
> + else
> + flags = VIR_DOMAIN_AFFECT_CONFIG;
> + }
> +
> + if (flags & VIR_DOMAIN_AFFECT_LIVE) {
> + if (!isActive) {
> + qemuReportError(VIR_ERR_OPERATION_INVALID,
> + "%s", _("domain is not running"));
> + goto cleanup;
> + }
> +
> + if (!qemuCgroupControllerActive(driver, VIR_CGROUP_CONTROLLER_CPUSET)) {
> + qemuReportError(VIR_ERR_OPERATION_INVALID,
> + "%s", _("cgroup cpuset controller is not mounted"));
> + goto cleanup;
> + }
> +
> + if (virCgroupForDomain(driver->cgroup, vm->def->name, &group, 0) != 0) {
> + qemuReportError(VIR_ERR_INTERNAL_ERROR,
> + _("cannot find cgroup for domain %s"), vm->def->name);
> + goto cleanup;
> + }
> + }
> +
> + if (flags & VIR_DOMAIN_AFFECT_CONFIG) {
> + if (!vm->persistent) {
> + qemuReportError(VIR_ERR_OPERATION_INVALID, "%s",
> + _("cannot change persistent config of a transient domain"));
> + goto cleanup;
> + }
> + if (!(persistentDef = virDomainObjGetPersistentDef(driver->caps, vm)))
> + goto cleanup;
> + }
> +
> + ret = 0;
> + for (i = 0; i < nparams; i++) {
> + virTypedParameterPtr param = ¶ms[i];
> +
> + if (STREQ(param->field, VIR_DOMAIN_NUMA_NODESET)) {
> + int rc;
> + if (param->type != VIR_TYPED_PARAM_STRING) {
> + qemuReportError(VIR_ERR_INVALID_ARG, "%s",
> + _("invalid type for numa nodeset tunable, expected a 'string'"));
> + ret = -1;
> + continue;
> + }
> +
> + if (flags & VIR_DOMAIN_AFFECT_LIVE) {
> + rc = virCgroupSetCpusetMems(group, params[i].value.s);
> + if (rc != 0) {
> + virReportSystemError(-rc, "%s",
> + _("unable to set memory hard_limit tunable"));
> + ret = -1;
> + continue;
> + }
> + }
> +
> + if (flags & VIR_DOMAIN_AFFECT_CONFIG) {
> + char *oldnodemask = strdup(persistentDef->numatune.memory.nodemask);
> + if (!oldnodemask) {
> + virReportOOMError();
> + ret = -1;
> + continue;
> + }
> + if (virDomainCpuSetParse((const char **)¶ms[i].value.s,
> + 0,
> + persistentDef->numatune.memory.nodemask,
> + VIR_DOMAIN_CPUMASK_LEN) < 0) {
> + VIR_FREE(persistentDef->numatune.memory.nodemask);
> + persistentDef->numatune.memory.nodemask = oldnodemask;
> + ret = -1;
> + continue;
> + }
> + }
> + } else if (STREQ(param->field, VIR_DOMAIN_NUMA_MODE)) {
> + int rc;
> + if (param->type != VIR_TYPED_PARAM_ULLONG) {
> + qemuReportError(VIR_ERR_INVALID_ARG, "%s",
> + _("invalid type for numa strict tunable, expected a 'ullong'"));
> + ret = -1;
> + continue;
> + }
> +
> + if (flags & VIR_DOMAIN_AFFECT_LIVE) {
> + switch(params[i].value.i) {
> + case VIR_DOMAIN_NUMATUNE_MEM_STRICT:
> + rc = virCgroupSetCpusetHardwall(group, params[i].value.i);
> + if (rc != 0) {
> + virReportSystemError(-rc, "%s",
> + _("unable to set memory hard_limit tunable"));
> + ret = -1;
> + }
> + break;
> + case VIR_DOMAIN_NUMATUNE_MEM_PREFERRED:
> + rc = virCgroupSetCpusetMemorySpreadPage(group, 0);
> + if (rc != 0) {
> + virReportSystemError(-rc,
> + _("Unable to set cpuset.memory_spread_page for domain %s"),
> + vm->def->name);
> + goto cleanup;
> + }
> + rc = virCgroupSetCpusetMemorySpreadSlab(group, 0);
> + if (rc != 0) {
> + virReportSystemError(-rc,
> + _("Unable to set cpuset.memory_spread_slab for domain %s"),
> + vm->def->name);
> + goto cleanup;
> + }
> + break;
> + case VIR_DOMAIN_NUMATUNE_MEM_INTERLEAVE:
> + rc = virCgroupSetCpusetMemorySpreadPage(group, 1);
> + if (rc != 0) {
> + virReportSystemError(-rc,
> + _("Unable to set cpuset.memory_spread_page for domain %s"),
> + vm->def->name);
> + goto cleanup;
> + }
> + rc = virCgroupSetCpusetMemorySpreadSlab(group, 1);
> + if (rc != 0) {
> + virReportSystemError(-rc,
> + _("Unable to set cpuset.memory_spread_slab for domain %s"),
> + vm->def->name);
> + goto cleanup;
> + }
> +
> + break;
> + default:
> + qemuReportError(VIR_ERR_INVALID_ARG, _("Unsupported mode"));
> + goto cleanup;
> + }
> + }
> +
> + if (flags & VIR_DOMAIN_AFFECT_CONFIG) {
> + persistentDef->numatune.memory.mode = params[i].value.i;
> + }
> + } else {
> + qemuReportError(VIR_ERR_INVALID_ARG,
> + _("Parameter `%s' not supported"), param->field);
> + ret = -1;
> + }
> + }
> +
> + if (flags & VIR_DOMAIN_AFFECT_CONFIG) {
> + if (virDomainSaveConfig(driver->configDir, persistentDef) < 0)
> + ret = -1;
> + }
> +
> +cleanup:
> + virCgroupFree(&group);
> + if (vm)
> + virDomainObjUnlock(vm);
> + qemuDriverUnlock(driver);
> + return ret;
> +}
> +
> +static int qemuDomainGetNumaParameters(virDomainPtr dom,
> + virTypedParameterPtr params,
> + int *nparams,
> + unsigned int flags)
> +{
> + struct qemud_driver *driver = dom->conn->privateData;
> + int i;
> + virCgroupPtr group = NULL;
> + virDomainObjPtr vm = NULL;
> + virDomainDefPtr persistentDef = NULL;
> + char *nodeset = NULL;
> + unsigned long long val;
> + int ret = -1;
> + int rc;
> + bool isActive;
> +
> + virCheckFlags(VIR_DOMAIN_AFFECT_LIVE |
> + VIR_DOMAIN_AFFECT_CONFIG |
> + VIR_TYPED_PARAM_STRING_OKAY, -1);
> +
> + qemuDriverLock(driver);
> +
> + flags &= ~VIR_TYPED_PARAM_STRING_OKAY;
> +
> + vm = virDomainFindByUUID(&driver->domains, dom->uuid);
> +
> + if (vm == NULL) {
> + qemuReportError(VIR_ERR_INTERNAL_ERROR,
> + _("No such domain %s"), dom->uuid);
> + goto cleanup;
> + }
> +
> + isActive = virDomainObjIsActive(vm);
> +
> + if (flags == VIR_DOMAIN_AFFECT_CURRENT) {
> + if (isActive)
> + flags = VIR_DOMAIN_AFFECT_LIVE;
> + else
> + flags = VIR_DOMAIN_AFFECT_CONFIG;
> + }
> +
> + if (flags & VIR_DOMAIN_AFFECT_LIVE) {
> + if (!isActive) {
> + qemuReportError(VIR_ERR_OPERATION_INVALID,
> + "%s", _("domain is not running"));
> + goto cleanup;
> + }
> +
> + if (!qemuCgroupControllerActive(driver, VIR_CGROUP_CONTROLLER_MEMORY)) {
> + qemuReportError(VIR_ERR_OPERATION_INVALID,
> + "%s", _("cgroup memory controller is not mounted"));
> + goto cleanup;
> + }
> +
> + if (virCgroupForDomain(driver->cgroup, vm->def->name, &group, 0) != 0) {
> + qemuReportError(VIR_ERR_INTERNAL_ERROR,
> + _("cannot find cgroup for domain %s"), vm->def->name);
> + goto cleanup;
> + }
> + }
> +
> + if (flags & VIR_DOMAIN_AFFECT_CONFIG) {
> + if (!vm->persistent) {
> + qemuReportError(VIR_ERR_OPERATION_INVALID, "%s",
> + _("cannot change persistent config of a transient domain"));
> + goto cleanup;
> + }
> + if (!(persistentDef = virDomainObjGetPersistentDef(driver->caps, vm)))
> + goto cleanup;
> + }
> +
> + if ((*nparams) == 0) {
> + *nparams = QEMU_NB_NUMA_PARAM;
> + ret = 0;
> + goto cleanup;
> + }
> +
> + if ((*nparams) < QEMU_NB_NUMA_PARAM) {
> + qemuReportError(VIR_ERR_INVALID_ARG,
> + "%s", _("Invalid parameter count"));
> + goto cleanup;
> + }
> +
> + if (flags & VIR_DOMAIN_AFFECT_CONFIG) {
> + for (i = 0; i < *nparams; i++) {
> + virMemoryParameterPtr param = ¶ms[i];
> + val = 0;
> + param->value.ul = 0;
> + param->type = VIR_TYPED_PARAM_ULLONG;
> +
> + switch (i) {
> + case 0: /* fill numa nodeset here */
> + if (virStrcpyStatic(param->field, VIR_DOMAIN_NUMA_NODESET) == NULL) {
> + qemuReportError(VIR_ERR_INTERNAL_ERROR,
> + "%s", _("Field numa nodeset too long for destination"));
> + goto cleanup;
> + }
> + if (persistentDef->numatune.memory.nodemask) {
> + nodeset = virDomainCpuSetFormat(persistentDef->numatune.memory.nodemask,
> + VIR_DOMAIN_CPUMASK_LEN);
> + if (!nodeset) {
> + qemuReportError(VIR_ERR_INTERNAL_ERROR,
> + "%s", _("failed to format nodeset for NUMA memory tuning"));
> + goto cleanup;
> + }
> + param->value.s = nodeset;
> + nodeset = NULL;
> + } else {
> + param->value.s = strdup("");
> + }
> + param->type = VIR_TYPED_PARAM_STRING;
> + break;
> +
> + case 1: /* fill numa mode here */
> + if (virStrcpyStatic(param->field, VIR_DOMAIN_NUMA_MODE) == NULL) {
> + qemuReportError(VIR_ERR_INTERNAL_ERROR,
> + "%s", _("Field numa mode too long for destination"));
> + goto cleanup;
> + }
> + param->value.i = persistentDef->numatune.memory.mode;
> + break;
> +
> + default:
> + break;
> + /* should not hit here */
> + }
> + }
> + goto out;
> + }
> +
> + for (i = 0; i < QEMU_NB_NUMA_PARAM; i++) {
> + virTypedParameterPtr param = ¶ms[i];
> + val = 0;
> + param->value.ul = 0;
> + param->type = VIR_TYPED_PARAM_ULLONG;
> +
> + /* Coverity does not realize that if we get here, group is set. */
> + sa_assert(group);
> +
> + switch (i) {
> + case 0: /* fill numa nodeset here */
> + rc = virCgroupGetCpusetHardwall(group, &val);
> + rc = virCgroupGetCpusetMems(group, &nodeset);
> + if (rc != 0) {
> + virReportSystemError(-rc, "%s",
> + _("unable to get numa nodeset"));
> + goto cleanup;
> + }
> + if (virStrcpyStatic(param->field, VIR_DOMAIN_NUMA_NODESET) == NULL) {
> + qemuReportError(VIR_ERR_INTERNAL_ERROR,
> + "%s", _("Field numa nodeset too long for destination"));
> + VIR_FREE(nodeset);
> + goto cleanup;
> + }
> + param->value.s = nodeset;
> + param->type = VIR_TYPED_PARAM_STRING;
> + break;
> +
> + case 1: /* file numa mode here */
> + if (virStrcpyStatic(param->field, VIR_DOMAIN_NUMA_MODE) == NULL) {
> + qemuReportError(VIR_ERR_INTERNAL_ERROR,
> + "%s", _("Field numa exclusive too long for destination"));
> + goto cleanup;
> + }
> + rc = virCgroupGetCpusetHardwall(group, &val);
> + if (rc != 0) {
> + virReportSystemError(-rc, "%s",
> + _("unable to get numa mode"));
> + goto cleanup;
> + }
> + if (val) {
> + param->value.ul = VIR_DOMAIN_NUMATUNE_MEM_STRICT;
> + break;
> + }
> + rc = virCgroupGetCpusetMemorySpreadPage(group, &val);
> + if (rc != 0) {
> + virReportSystemError(-rc, "%s",
> + _("unable to get numa mode"));
> + goto cleanup;
> + }
> + if (val) {
> + param->value.ul = VIR_DOMAIN_NUMATUNE_MEM_INTERLEAVE;
> + break;
> + } else {
> + param->value.ul = VIR_DOMAIN_NUMATUNE_MEM_PREFERRED;
> + break;
> + }
> + param->value.ul = VIR_DOMAIN_NUMATUNE_MEM_NONE;
> + break;
> +
> + default:
> + break;
> + /* should not hit here */
> + }
> + }
> +
> +out:
> + *nparams = QEMU_NB_NUMA_PARAM;
> + ret = 0;
> +
> +cleanup:
> + if (group)
> + virCgroupFree(&group);
> + if (vm)
> + virDomainObjUnlock(vm);
> + qemuDriverUnlock(driver);
> + return ret;
> +}
Same API comments as with patch 7 of course. In terms of logic, we can
only allow the nodeset mask to be changed on the fly given current cpuset
functionality. We can't support switching between policy modes.
Daniel
--
|: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org -o- http://virt-manager.org :|
|: http://autobuild.org -o- http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :|
--
libvir-list mailing list
libvir-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/libvir-list