static int
qemuSetupCpuCgroup(virQEMUDriverPtr driver,
virDomainObjPtr vm)
{
qemuDomainObjPrivatePtr priv = vm->privateData;
virObjectEventPtr event = NULL;
virTypedParameterPtr eventParams = NULL;
int eventNparams = 0;
int eventMaxparams = 0;
if (!virCgroupHasController(priv->cgroup, VIR_CGROUP_CONTROLLER_CPU)) {
if (vm->def->cputune.sharesSpecified) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
_("CPU tuning is not available on this host"));
return -1;
} else {
return 0;
}
}
if (vm->def->cputune.sharesSpecified) {
unsigned long long val;
if (virCgroupSetCpuShares(priv->cgroup, vm->def->cputune.shares) < 0)
return -1;
if (virCgroupGetCpuShares(priv->cgroup, &val) < 0)
return -1;
if (vm->def->cputune.shares != val) {
vm->def->cputune.shares = val;
if (virTypedParamsAddULLong(&eventParams, &eventNparams,
&eventMaxparams,
VIR_DOMAIN_TUNABLE_CPU_CPU_SHARES,
val) < 0)
return -1;
event = virDomainEventTunableNewFromObj(vm, eventParams, eventNparams);
}
if (event)
qemuDomainEventQueue(driver, event);
}
return 0;
}
/** qemuMigrationSetOffline
* Pause domain for non-live migration.
*/
int
qemuMigrationSetOffline(struct qemud_driver *driver,
virDomainObjPtr vm)
{
int ret;
ret = qemuProcessStopCPUs(driver, vm);
if (ret == 0) {
virDomainEventPtr event;
event = virDomainEventNewFromObj(vm,
VIR_DOMAIN_EVENT_SUSPENDED,
VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED);
if (event)
qemuDomainEventQueue(driver, event);
}
return ret;
}
/**
* qemuBlockJobEventProcess:
* @driver: qemu driver
* @vm: domain
* @disk: domain disk
* @type: block job type
* @status: block job status
*
* Update disk's mirror state in response to a block job event
* from QEMU. For mirror state's that must survive libvirt
* restart, also update the domain's status XML.
*/
void
qemuBlockJobEventProcess(virQEMUDriverPtr driver,
virDomainObjPtr vm,
virDomainDiskDefPtr disk,
qemuDomainAsyncJob asyncJob,
int type,
int status)
{
virObjectEventPtr event = NULL;
virObjectEventPtr event2 = NULL;
const char *path;
virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
virDomainDiskDefPtr persistDisk = NULL;
qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
VIR_DEBUG("disk=%s, mirrorState=%s, type=%d, status=%d",
disk->dst,
NULLSTR(virDomainDiskMirrorStateTypeToString(disk->mirrorState)),
type,
status);
/* Have to generate two variants of the event for old vs. new
* client callbacks */
if (type == VIR_DOMAIN_BLOCK_JOB_TYPE_COMMIT &&
disk->mirrorJob == VIR_DOMAIN_BLOCK_JOB_TYPE_ACTIVE_COMMIT)
type = disk->mirrorJob;
path = virDomainDiskGetSource(disk);
event = virDomainEventBlockJobNewFromObj(vm, path, type, status);
event2 = virDomainEventBlockJob2NewFromObj(vm, disk->dst, type, status);
/* If we completed a block pull or commit, then update the XML
* to match. */
switch ((virConnectDomainEventBlockJobStatus) status) {
case VIR_DOMAIN_BLOCK_JOB_COMPLETED:
if (disk->mirrorState == VIR_DOMAIN_DISK_MIRROR_STATE_PIVOT) {
if (vm->newDef) {
virStorageSourcePtr copy = NULL;
if ((persistDisk = virDomainDiskByName(vm->newDef,
disk->dst, false))) {
copy = virStorageSourceCopy(disk->mirror, false);
if (!copy ||
virStorageSourceInitChainElement(copy,
persistDisk->src,
true) < 0) {
VIR_WARN("Unable to update persistent definition "
"on vm %s after block job",
vm->def->name);
virStorageSourceFree(copy);
copy = NULL;
persistDisk = NULL;
}
}
if (copy) {
virStorageSourceFree(persistDisk->src);
persistDisk->src = copy;
}
}
/* XXX We want to revoke security labels as well as audit that
* revocation, before dropping the original source. But it gets
* tricky if both source and mirror share common backing files (we
* want to only revoke the non-shared portion of the chain); so for
* now, we leak the access to the original. */
virDomainLockImageDetach(driver->lockManager, vm, disk->src);
virStorageSourceFree(disk->src);
disk->src = disk->mirror;
} else {
if (disk->mirror) {
virDomainLockImageDetach(driver->lockManager, vm, disk->mirror);
virStorageSourceFree(disk->mirror);
}
}
/* Recompute the cached backing chain to match our
* updates. Better would be storing the chain ourselves
* rather than reprobing, but we haven't quite completed
* that conversion to use our XML tracking. */
disk->mirror = NULL;
disk->mirrorState = VIR_DOMAIN_DISK_MIRROR_STATE_NONE;
disk->mirrorJob = VIR_DOMAIN_BLOCK_JOB_TYPE_UNKNOWN;
disk->src->id = 0;
ignore_value(qemuDomainDetermineDiskChain(driver, vm, disk,
true, true));
ignore_value(qemuBlockNodeNamesDetect(driver, vm, asyncJob));
diskPriv->blockjob = false;
break;
case VIR_DOMAIN_BLOCK_JOB_READY:
disk->mirrorState = VIR_DOMAIN_DISK_MIRROR_STATE_READY;
break;
case VIR_DOMAIN_BLOCK_JOB_FAILED:
case VIR_DOMAIN_BLOCK_JOB_CANCELED:
if (disk->mirror) {
virDomainLockImageDetach(driver->lockManager, vm, disk->mirror);
virStorageSourceFree(disk->mirror);
disk->mirror = NULL;
}
disk->mirrorState = VIR_DOMAIN_DISK_MIRROR_STATE_NONE;
disk->mirrorJob = VIR_DOMAIN_BLOCK_JOB_TYPE_UNKNOWN;
diskPriv->blockjob = false;
break;
case VIR_DOMAIN_BLOCK_JOB_LAST:
break;
}
if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
VIR_WARN("Unable to save status on vm %s after block job", vm->def->name);
if (status == VIR_DOMAIN_BLOCK_JOB_COMPLETED && vm->newDef) {
if (virDomainSaveConfig(cfg->configDir, driver->caps, vm->newDef) < 0)
VIR_WARN("Unable to update persistent definition on vm %s "
"after block job", vm->def->name);
}
qemuDomainEventQueue(driver, event);
qemuDomainEventQueue(driver, event2);
virObjectUnref(cfg);
}
int
qemuMigrationWaitForCompletion(struct qemud_driver *driver, virDomainObjPtr vm)
{
int ret = -1;
int status;
unsigned long long memProcessed;
unsigned long long memRemaining;
unsigned long long memTotal;
qemuDomainObjPrivatePtr priv = vm->privateData;
priv->jobInfo.type = VIR_DOMAIN_JOB_UNBOUNDED;
while (priv->jobInfo.type == VIR_DOMAIN_JOB_UNBOUNDED) {
/* Poll every 50ms for progress & to allow cancellation */
struct timespec ts = { .tv_sec = 0, .tv_nsec = 50 * 1000 * 1000ull };
struct timeval now;
int rc;
const char *job;
switch (priv->jobActive) {
case QEMU_JOB_MIGRATION_OUT:
job = _("migration job");
break;
case QEMU_JOB_SAVE:
job = _("domain save job");
break;
case QEMU_JOB_DUMP:
job = _("domain core dump job");
break;
default:
job = _("job");
}
if (!virDomainObjIsActive(vm)) {
qemuReportError(VIR_ERR_INTERNAL_ERROR, _("%s: %s"),
job, _("guest unexpectedly quit"));
goto cleanup;
}
if (priv->jobSignals & QEMU_JOB_SIGNAL_CANCEL) {
priv->jobSignals ^= QEMU_JOB_SIGNAL_CANCEL;
VIR_DEBUG0("Cancelling job at client request");
qemuDomainObjEnterMonitorWithDriver(driver, vm);
rc = qemuMonitorMigrateCancel(priv->mon);
qemuDomainObjExitMonitorWithDriver(driver, vm);
if (rc < 0) {
VIR_WARN0("Unable to cancel job");
}
} else if (priv->jobSignals & QEMU_JOB_SIGNAL_SUSPEND) {
priv->jobSignals ^= QEMU_JOB_SIGNAL_SUSPEND;
VIR_DEBUG0("Pausing domain for non-live migration");
if (qemuMigrationSetOffline(driver, vm) < 0)
VIR_WARN0("Unable to pause domain");
} else if (priv->jobSignals & QEMU_JOB_SIGNAL_MIGRATE_DOWNTIME) {
unsigned long long ms = priv->jobSignalsData.migrateDowntime;
priv->jobSignals ^= QEMU_JOB_SIGNAL_MIGRATE_DOWNTIME;
priv->jobSignalsData.migrateDowntime = 0;
VIR_DEBUG("Setting migration downtime to %llums", ms);
qemuDomainObjEnterMonitorWithDriver(driver, vm);
rc = qemuMonitorSetMigrationDowntime(priv->mon, ms);
qemuDomainObjExitMonitorWithDriver(driver, vm);
if (rc < 0)
VIR_WARN0("Unable to set migration downtime");
} else if (priv->jobSignals & QEMU_JOB_SIGNAL_MIGRATE_SPEED) {
unsigned long bandwidth = priv->jobSignalsData.migrateBandwidth;
priv->jobSignals ^= QEMU_JOB_SIGNAL_MIGRATE_SPEED;
priv->jobSignalsData.migrateBandwidth = 0;
VIR_DEBUG("Setting migration bandwidth to %luMbs", bandwidth);
qemuDomainObjEnterMonitorWithDriver(driver, vm);
rc = qemuMonitorSetMigrationSpeed(priv->mon, bandwidth);
qemuDomainObjExitMonitorWithDriver(driver, vm);
if (rc < 0)
VIR_WARN0("Unable to set migration speed");
}
/* Repeat check because the job signals might have caused
* guest to die
*/
if (!virDomainObjIsActive(vm)) {
qemuReportError(VIR_ERR_INTERNAL_ERROR, _("%s: %s"),
job, _("guest unexpectedly quit"));
goto cleanup;
}
qemuDomainObjEnterMonitorWithDriver(driver, vm);
rc = qemuMonitorGetMigrationStatus(priv->mon,
&status,
&memProcessed,
&memRemaining,
&memTotal);
qemuDomainObjExitMonitorWithDriver(driver, vm);
if (rc < 0) {
priv->jobInfo.type = VIR_DOMAIN_JOB_FAILED;
goto cleanup;
}
if (gettimeofday(&now, NULL) < 0) {
priv->jobInfo.type = VIR_DOMAIN_JOB_FAILED;
virReportSystemError(errno, "%s",
_("cannot get time of day"));
goto cleanup;
}
priv->jobInfo.timeElapsed = timeval_to_ms(now) - priv->jobStart;
switch (status) {
case QEMU_MONITOR_MIGRATION_STATUS_INACTIVE:
priv->jobInfo.type = VIR_DOMAIN_JOB_NONE;
qemuReportError(VIR_ERR_OPERATION_FAILED,
_("%s: %s"), job, _("is not active"));
break;
case QEMU_MONITOR_MIGRATION_STATUS_ACTIVE:
priv->jobInfo.dataTotal = memTotal;
priv->jobInfo.dataRemaining = memRemaining;
priv->jobInfo.dataProcessed = memProcessed;
priv->jobInfo.memTotal = memTotal;
priv->jobInfo.memRemaining = memRemaining;
priv->jobInfo.memProcessed = memProcessed;
break;
case QEMU_MONITOR_MIGRATION_STATUS_COMPLETED:
priv->jobInfo.type = VIR_DOMAIN_JOB_COMPLETED;
ret = 0;
break;
case QEMU_MONITOR_MIGRATION_STATUS_ERROR:
priv->jobInfo.type = VIR_DOMAIN_JOB_FAILED;
qemuReportError(VIR_ERR_OPERATION_FAILED,
_("%s: %s"), job, _("unexpectedly failed"));
break;
case QEMU_MONITOR_MIGRATION_STATUS_CANCELLED:
priv->jobInfo.type = VIR_DOMAIN_JOB_CANCELLED;
qemuReportError(VIR_ERR_OPERATION_FAILED,
_("%s: %s"), job, _("canceled by client"));
break;
}
virDomainObjUnlock(vm);
qemuDriverUnlock(driver);
nanosleep(&ts, NULL);
qemuDriverLock(driver);
virDomainObjLock(vm);
}
cleanup:
return ret;
}
/* Prepare is the first step, and it runs on the destination host.
*
* This version starts an empty VM listening on a localhost TCP port, and
* sets up the corresponding virStream to handle the incoming data.
*/
int
qemuMigrationPrepareTunnel(struct qemud_driver *driver,
virConnectPtr dconn,
virStreamPtr st,
const char *dname,
const char *dom_xml)
{
virDomainDefPtr def = NULL;
virDomainObjPtr vm = NULL;
virDomainEventPtr event = NULL;
int ret = -1;
int internalret;
int dataFD[2] = { -1, -1 };
virBitmapPtr qemuCaps = NULL;
qemuDomainObjPrivatePtr priv = NULL;
struct timeval now;
if (gettimeofday(&now, NULL) < 0) {
virReportSystemError(errno, "%s",
_("cannot get time of day"));
return -1;
}
/* Parse the domain XML. */
if (!(def = virDomainDefParseString(driver->caps, dom_xml,
VIR_DOMAIN_XML_INACTIVE)))
goto cleanup;
if (!qemuMigrationIsAllowed(def))
goto cleanup;
/* Target domain name, maybe renamed. */
if (dname) {
VIR_FREE(def->name);
def->name = strdup(dname);
if (def->name == NULL)
goto cleanup;
}
if (virDomainObjIsDuplicate(&driver->domains, def, 1) < 0)
goto cleanup;
if (!(vm = virDomainAssignDef(driver->caps,
&driver->domains,
def, true))) {
/* virDomainAssignDef already set the error */
goto cleanup;
}
def = NULL;
priv = vm->privateData;
if (qemuDomainObjBeginJobWithDriver(driver, vm) < 0)
goto cleanup;
priv->jobActive = QEMU_JOB_MIGRATION_OUT;
/* Domain starts inactive, even if the domain XML had an id field. */
vm->def->id = -1;
if (pipe(dataFD) < 0 ||
virSetCloseExec(dataFD[0]) < 0) {
virReportSystemError(errno, "%s",
_("cannot create pipe for tunnelled migration"));
goto endjob;
}
/* check that this qemu version supports the interactive exec */
if (qemuCapsExtractVersionInfo(vm->def->emulator, vm->def->os.arch,
NULL, &qemuCaps) < 0) {
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("Cannot determine QEMU argv syntax %s"),
vm->def->emulator);
goto endjob;
}
/* Start the QEMU daemon, with the same command-line arguments plus
* -incoming stdin (which qemu_command might convert to exec:cat or fd:n)
*/
internalret = qemuProcessStart(dconn, driver, vm, "stdin", true, dataFD[1],
NULL, VIR_VM_OP_MIGRATE_IN_START);
if (internalret < 0) {
qemuAuditDomainStart(vm, "migrated", false);
/* Note that we don't set an error here because qemuProcessStart
* should have already done that.
*/
if (!vm->persistent) {
virDomainRemoveInactive(&driver->domains, vm);
vm = NULL;
}
goto endjob;
}
if (virFDStreamOpen(st, dataFD[0]) < 0) {
qemuAuditDomainStart(vm, "migrated", false);
qemuProcessStop(driver, vm, 0);
if (!vm->persistent) {
if (qemuDomainObjEndJob(vm) > 0)
virDomainRemoveInactive(&driver->domains, vm);
vm = NULL;
}
virReportSystemError(errno, "%s",
_("cannot pass pipe for tunnelled migration"));
goto endjob;
}
qemuAuditDomainStart(vm, "migrated", true);
event = virDomainEventNewFromObj(vm,
VIR_DOMAIN_EVENT_STARTED,
VIR_DOMAIN_EVENT_STARTED_MIGRATED);
ret = 0;
endjob:
if (vm &&
qemuDomainObjEndJob(vm) == 0)
vm = NULL;
/* We set a fake job active which is held across
* API calls until the finish() call. This prevents
* any other APIs being invoked while incoming
* migration is taking place
*/
if (vm &&
virDomainObjIsActive(vm)) {
priv->jobActive = QEMU_JOB_MIGRATION_IN;
priv->jobInfo.type = VIR_DOMAIN_JOB_UNBOUNDED;
priv->jobStart = timeval_to_ms(now);
}
cleanup:
qemuCapsFree(qemuCaps);
virDomainDefFree(def);
VIR_FORCE_CLOSE(dataFD[0]);
VIR_FORCE_CLOSE(dataFD[1]);
if (vm)
virDomainObjUnlock(vm);
if (event)
qemuDomainEventQueue(driver, event);
qemuDriverUnlock(driver);
return ret;
}
int
qemuMigrationPrepareDirect(struct qemud_driver *driver,
virConnectPtr dconn,
const char *uri_in,
char **uri_out,
const char *dname,
const char *dom_xml)
{
static int port = 0;
virDomainDefPtr def = NULL;
virDomainObjPtr vm = NULL;
int this_port;
char *hostname = NULL;
char migrateFrom [64];
const char *p;
virDomainEventPtr event = NULL;
int ret = -1;
int internalret;
qemuDomainObjPrivatePtr priv = NULL;
struct timeval now;
if (gettimeofday(&now, NULL) < 0) {
virReportSystemError(errno, "%s",
_("cannot get time of day"));
return -1;
}
/* The URI passed in may be NULL or a string "tcp://somehostname:port".
*
* If the URI passed in is NULL then we allocate a port number
* from our pool of port numbers and return a URI of
* "tcp://ourhostname:port".
*
* If the URI passed in is not NULL then we try to parse out the
* port number and use that (note that the hostname is assumed
* to be a correct hostname which refers to the target machine).
*/
if (uri_in == NULL) {
this_port = QEMUD_MIGRATION_FIRST_PORT + port++;
if (port == QEMUD_MIGRATION_NUM_PORTS) port = 0;
/* Get hostname */
if ((hostname = virGetHostname(NULL)) == NULL)
goto cleanup;
if (STRPREFIX(hostname, "localhost")) {
qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("hostname on destination resolved to localhost, but migration requires an FQDN"));
goto cleanup;
}
/* XXX this really should have been a properly well-formed
* URI, but we can't add in tcp:// now without breaking
* compatability with old targets. We at least make the
* new targets accept both syntaxes though.
*/
/* Caller frees */
internalret = virAsprintf(uri_out, "tcp:%s:%d", hostname, this_port);
if (internalret < 0) {
virReportOOMError();
goto cleanup;
}
} else {
/* Check the URI starts with "tcp:". We will escape the
* URI when passing it to the qemu monitor, so bad
* characters in hostname part don't matter.
*/
if (!STRPREFIX (uri_in, "tcp:")) {
qemuReportError (VIR_ERR_INVALID_ARG,
"%s", _("only tcp URIs are supported for KVM/QEMU migrations"));
goto cleanup;
}
/* Get the port number. */
p = strrchr (uri_in, ':');
if (p == strchr(uri_in, ':')) {
/* Generate a port */
this_port = QEMUD_MIGRATION_FIRST_PORT + port++;
if (port == QEMUD_MIGRATION_NUM_PORTS)
port = 0;
/* Caller frees */
if (virAsprintf(uri_out, "%s:%d", uri_in, this_port) < 0) {
virReportOOMError();
goto cleanup;
}
} else {
p++; /* definitely has a ':' in it, see above */
this_port = virParseNumber (&p);
if (this_port == -1 || p-uri_in != strlen (uri_in)) {
qemuReportError(VIR_ERR_INVALID_ARG,
"%s", _("URI ended with incorrect ':port'"));
goto cleanup;
}
}
}
if (*uri_out)
VIR_DEBUG("Generated uri_out=%s", *uri_out);
/* Parse the domain XML. */
if (!(def = virDomainDefParseString(driver->caps, dom_xml,
VIR_DOMAIN_XML_INACTIVE)))
goto cleanup;
if (!qemuMigrationIsAllowed(def))
goto cleanup;
/* Target domain name, maybe renamed. */
if (dname) {
VIR_FREE(def->name);
def->name = strdup(dname);
if (def->name == NULL)
goto cleanup;
}
if (virDomainObjIsDuplicate(&driver->domains, def, 1) < 0)
goto cleanup;
if (!(vm = virDomainAssignDef(driver->caps,
&driver->domains,
def, true))) {
/* virDomainAssignDef already set the error */
goto cleanup;
}
def = NULL;
priv = vm->privateData;
if (qemuDomainObjBeginJobWithDriver(driver, vm) < 0)
goto cleanup;
priv->jobActive = QEMU_JOB_MIGRATION_OUT;
/* Domain starts inactive, even if the domain XML had an id field. */
vm->def->id = -1;
/* Start the QEMU daemon, with the same command-line arguments plus
* -incoming tcp:0.0.0.0:port
*/
snprintf (migrateFrom, sizeof (migrateFrom), "tcp:0.0.0.0:%d", this_port);
if (qemuProcessStart(dconn, driver, vm, migrateFrom, true,
-1, NULL, VIR_VM_OP_MIGRATE_IN_START) < 0) {
qemuAuditDomainStart(vm, "migrated", false);
/* Note that we don't set an error here because qemuProcessStart
* should have already done that.
*/
if (!vm->persistent) {
if (qemuDomainObjEndJob(vm) > 0)
virDomainRemoveInactive(&driver->domains, vm);
vm = NULL;
}
goto endjob;
}
qemuAuditDomainStart(vm, "migrated", true);
event = virDomainEventNewFromObj(vm,
VIR_DOMAIN_EVENT_STARTED,
VIR_DOMAIN_EVENT_STARTED_MIGRATED);
ret = 0;
endjob:
if (vm &&
qemuDomainObjEndJob(vm) == 0)
vm = NULL;
/* We set a fake job active which is held across
* API calls until the finish() call. This prevents
* any other APIs being invoked while incoming
* migration is taking place
*/
if (vm &&
virDomainObjIsActive(vm)) {
priv->jobActive = QEMU_JOB_MIGRATION_IN;
priv->jobInfo.type = VIR_DOMAIN_JOB_UNBOUNDED;
priv->jobStart = timeval_to_ms(now);
}
cleanup:
VIR_FREE(hostname);
virDomainDefFree(def);
if (ret != 0)
VIR_FREE(*uri_out);
if (vm)
virDomainObjUnlock(vm);
if (event)
qemuDomainEventQueue(driver, event);
return ret;
}