/**
* qemuAgentGuestSync:
* @mon: Monitor
*
* Send guest-sync with unique ID
* and wait for reply. If we get one, check if
* received ID is equal to given.
*
* Returns: 0 on success,
* -1 otherwise
*/
static int
qemuAgentGuestSync(qemuAgentPtr mon)
{
int ret = -1;
int send_ret;
unsigned long long id, id_ret;
qemuAgentMessage sync_msg;
memset(&sync_msg, 0, sizeof(sync_msg));
if (virTimeMillisNow(&id) < 0)
return -1;
if (virAsprintf(&sync_msg.txBuffer,
"{\"execute\":\"guest-sync\", "
"\"arguments\":{\"id\":%llu}}", id) < 0) {
virReportOOMError();
return -1;
}
sync_msg.txLength = strlen(sync_msg.txBuffer);
VIR_DEBUG("Sending guest-sync command with ID: %llu", id);
send_ret = qemuAgentSend(mon, &sync_msg, true);
VIR_DEBUG("qemuAgentSend returned: %d", send_ret);
if (send_ret < 0) {
/* error reported */
goto cleanup;
}
if (!sync_msg.rxObject) {
qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("Missing monitor reply object"));
goto cleanup;
}
if (virJSONValueObjectGetNumberUlong(sync_msg.rxObject,
"return", &id_ret) < 0) {
qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("Malformed return value"));
goto cleanup;
}
VIR_DEBUG("Guest returned ID: %llu", id_ret);
if (id_ret != id) {
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("Guest agent returned ID: %llu instead of %llu"),
id_ret, id);
goto cleanup;
}
ret = 0;
cleanup:
virJSONValueFree(sync_msg.rxObject);
VIR_FREE(sync_msg.txBuffer);
return ret;
}
static int
qemuAgentIOProcessLine(qemuAgentPtr mon,
const char *line,
qemuAgentMessagePtr msg)
{
virJSONValuePtr obj = NULL;
int ret = -1;
unsigned long long id;
VIR_DEBUG("Line [%s]", line);
if (!(obj = virJSONValueFromString(line)))
goto cleanup;
if (obj->type != VIR_JSON_TYPE_OBJECT) {
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("Parsed JSON reply '%s' isn't an object"), line);
goto cleanup;
}
if (virJSONValueObjectHasKey(obj, "QMP") == 1) {
ret = 0;
} else if (virJSONValueObjectHasKey(obj, "event") == 1) {
ret = qemuAgentIOProcessEvent(mon, obj);
} else if (virJSONValueObjectHasKey(obj, "error") == 1 ||
virJSONValueObjectHasKey(obj, "return") == 1) {
if (msg) {
msg->rxObject = obj;
msg->finished = 1;
obj = NULL;
ret = 0;
} else {
/* If we've received something like:
* {"return": 1234}
* it is likely that somebody started GA
* which is now processing our previous
* guest-sync commands. Check if this is
* the case and don't report an error but
* return silently.
*/
if (virJSONValueObjectGetNumberUlong(obj, "return", &id) == 0) {
VIR_DEBUG("Ignoring delayed reply to guest-sync: %llu", id);
ret = 0;
goto cleanup;
}
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("Unexpected JSON reply '%s'"), line);
}
} else {
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("Unknown JSON reply '%s'"), line);
}
cleanup:
virJSONValueFree(obj);
return ret;
}
/**
* qemuAgentSend:
* @mon: Monitor
* @msg: Message
* @timeout: use timeout?
*
* Send @msg to agent @mon.
* Wait max QEMU_AGENT_WAIT_TIME for agent
* to reply.
*
* Returns: 0 on success,
* -2 on timeout,
* -1 otherwise
*/
static int qemuAgentSend(qemuAgentPtr mon,
qemuAgentMessagePtr msg,
bool timeout)
{
int ret = -1;
unsigned long long now, then = 0;
/* Check whether qemu quit unexpectedly */
if (mon->lastError.code != VIR_ERR_OK) {
VIR_DEBUG("Attempt to send command while error is set %s",
NULLSTR(mon->lastError.message));
virSetError(&mon->lastError);
return -1;
}
if (timeout) {
if (virTimeMillisNow(&now) < 0)
return -1;
then = now + QEMU_AGENT_WAIT_TIME;
}
mon->msg = msg;
qemuAgentUpdateWatch(mon);
while (!mon->msg->finished) {
if ((timeout && virCondWaitUntil(&mon->notify, &mon->lock, then) < 0) ||
(!timeout && virCondWait(&mon->notify, &mon->lock) < 0)) {
if (errno == ETIMEDOUT) {
qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("Guest agent not available for now"));
ret = -2;
} else {
qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("Unable to wait on monitor condition"));
}
goto cleanup;
}
}
if (mon->lastError.code != VIR_ERR_OK) {
VIR_DEBUG("Send command resulted in error %s",
NULLSTR(mon->lastError.message));
virSetError(&mon->lastError);
goto cleanup;
}
ret = 0;
cleanup:
mon->msg = NULL;
qemuAgentUpdateWatch(mon);
return ret;
}
/*
* qemuAgentFSThaw:
* @mon: Agent
*
* Issue guest-fsfreeze-thaw command to guest agent,
* which unfreezes all mounted file systems and returns
* number of thawed file systems on success.
*
* Returns: number of file system thawed on success,
* -1 on error.
*/
int qemuAgentFSThaw(qemuAgentPtr mon)
{
int ret = -1;
virJSONValuePtr cmd;
virJSONValuePtr reply = NULL;
cmd = qemuAgentMakeCommand("guest-fsfreeze-thaw", NULL);
if (!cmd)
return -1;
if (qemuAgentCommand(mon, cmd, &reply) < 0 ||
qemuAgentCheckError(cmd, reply) < 0)
goto cleanup;
if (virJSONValueObjectGetNumberInt(reply, "return", &ret) < 0) {
qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("malformed return value"));
}
cleanup:
virJSONValueFree(cmd);
virJSONValueFree(reply);
return ret;
}
int
qemuDomainHostdevNetConfigRestore(virDomainHostdevDefPtr hostdev,
char *stateDir)
{
char *linkdev = NULL;
virNetDevVPortProfilePtr virtPort;
int ret = -1;
int vf = -1;
int port_profile_associate = 0;
int isvf;
isvf = qemuDomainHostdevIsVirtualFunction(hostdev);
if (isvf <= 0) {
qemuReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("Interface type hostdev is currently supported on"
" SR-IOV Virtual Functions only"));
return ret;
}
if (qemuDomainHostdevNetDevice(hostdev, &linkdev, &vf) < 0)
return ret;
virtPort = virDomainNetGetActualVirtPortProfile(
hostdev->parent.data.net);
if (virtPort)
ret = qemuDomainHostdevNetConfigVirtPortProfile(linkdev, vf, virtPort,
hostdev->parent.data.net->mac, NULL,
port_profile_associate);
else
ret = virNetDevRestoreNetConfig(linkdev, vf, stateDir);
VIR_FREE(linkdev);
return ret;
}
static int
qemuAgentCheckError(virJSONValuePtr cmd,
virJSONValuePtr reply)
{
if (virJSONValueObjectHasKey(reply, "error")) {
virJSONValuePtr error = virJSONValueObjectGet(reply, "error");
char *cmdstr = virJSONValueToString(cmd);
char *replystr = virJSONValueToString(reply);
/* Log the full JSON formatted command & error */
VIR_DEBUG("unable to execute QEMU command %s: %s",
cmdstr, replystr);
/* Only send the user the command name + friendly error */
if (!error)
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("unable to execute QEMU command '%s'"),
qemuAgentCommandName(cmd));
else
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("unable to execute QEMU command '%s': %s"),
qemuAgentCommandName(cmd),
qemuAgentStringifyError(error));
VIR_FREE(cmdstr);
VIR_FREE(replystr);
return -1;
} else if (!virJSONValueObjectHasKey(reply, "return")) {
char *cmdstr = virJSONValueToString(cmd);
char *replystr = virJSONValueToString(reply);
VIR_DEBUG("Neither 'return' nor 'error' is set in the JSON reply %s: %s",
cmdstr, replystr);
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("unable to execute QEMU command '%s'"),
qemuAgentCommandName(cmd));
VIR_FREE(cmdstr);
VIR_FREE(replystr);
return -1;
}
return 0;
}
bool
qemuMigrationIsAllowed(virDomainDefPtr def)
{
if (def->nhostdevs > 0) {
qemuReportError(VIR_ERR_OPERATION_INVALID,
"%s", _("Domain with assigned host devices cannot be migrated"));
return false;
}
return true;
}
int
qemuPrepareHostdevUSBDevices(struct qemud_driver *driver,
const char *name,
usbDeviceList *list)
{
int i;
unsigned int count;
usbDevice *tmp;
count = usbDeviceListCount(list);
for (i = 0; i < count; i++) {
usbDevice *usb = usbDeviceListGet(list, i);
if ((tmp = usbDeviceListFind(driver->activeUsbHostdevs, usb))) {
const char *other_name = usbDeviceGetUsedBy(tmp);
if (other_name)
qemuReportError(VIR_ERR_OPERATION_INVALID,
_("USB device %s is in use by domain %s"),
usbDeviceGetName(tmp), other_name);
else
qemuReportError(VIR_ERR_OPERATION_INVALID,
_("USB device %s is already in use"),
usbDeviceGetName(tmp));
return -1;
}
usbDeviceSetUsedBy(usb, name);
VIR_DEBUG("Adding %03d.%03d dom=%s to activeUsbHostdevs",
usbDeviceGetBus(usb), usbDeviceGetDevno(usb), name);
/*
* The caller is responsible to steal these usb devices
* from the usbDeviceList that passed in on success,
* perform rollback on failure.
*/
if (usbDeviceListAdd(driver->activeUsbHostdevs, usb) < 0)
return -1;
}
return 0;
}
static int
qemuAgentCommand(qemuAgentPtr mon,
virJSONValuePtr cmd,
virJSONValuePtr *reply)
{
int ret = -1;
qemuAgentMessage msg;
char *cmdstr = NULL;
*reply = NULL;
if (qemuAgentGuestSync(mon) < 0) {
/* helper reported the error */
return -1;
}
memset(&msg, 0, sizeof(msg));
if (!(cmdstr = virJSONValueToString(cmd))) {
virReportOOMError();
goto cleanup;
}
if (virAsprintf(&msg.txBuffer, "%s" LINE_ENDING, cmdstr) < 0) {
virReportOOMError();
goto cleanup;
}
msg.txLength = strlen(msg.txBuffer);
VIR_DEBUG("Send command '%s' for write", cmdstr);
ret = qemuAgentSend(mon, &msg, false);
VIR_DEBUG("Receive command reply ret=%d rxObject=%p",
ret, msg.rxObject);
if (ret == 0) {
if (!msg.rxObject) {
qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("Missing monitor reply object"));
ret = -1;
} else {
*reply = msg.rxObject;
}
}
cleanup:
VIR_FREE(cmdstr);
VIR_FREE(msg.txBuffer);
return ret;
}
static int
qemuAgentOpenPty(const char *monitor)
{
int monfd;
if ((monfd = open(monitor, O_RDWR | O_NONBLOCK)) < 0) {
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("Unable to open monitor path %s"), monitor);
return -1;
}
if (virSetCloseExec(monfd) < 0) {
qemuReportError(VIR_ERR_INTERNAL_ERROR,
"%s", _("Unable to set monitor close-on-exec flag"));
goto error;
}
return monfd;
error:
VIR_FORCE_CLOSE(monfd);
return -1;
}
static int
qemuDomainHostdevNetConfigVirtPortProfile(const char *linkdev, int vf,
virNetDevVPortProfilePtr virtPort,
const unsigned char *macaddr,
const unsigned char *uuid,
int associate)
{
int ret = -1;
if (!virtPort)
return ret;
switch(virtPort->virtPortType) {
case VIR_NETDEV_VPORT_PROFILE_NONE:
case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
case VIR_NETDEV_VPORT_PROFILE_8021QBG:
case VIR_NETDEV_VPORT_PROFILE_LAST:
qemuReportError(VIR_ERR_CONFIG_UNSUPPORTED, _("virtualport type %s is "
"currently not supported on interfaces of type "
"hostdev"),
virNetDevVPortTypeToString(virtPort->virtPortType));
break;
case VIR_NETDEV_VPORT_PROFILE_8021QBH:
if (associate)
ret = virNetDevVPortProfileAssociate(NULL, virtPort, macaddr,
linkdev, vf, uuid,
VIR_NETDEV_VPORT_PROFILE_OP_CREATE, false);
else
ret = virNetDevVPortProfileDisassociate(NULL, virtPort,
macaddr, linkdev, vf,
VIR_NETDEV_VPORT_PROFILE_OP_DESTROY);
break;
}
return ret;
}
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;
}
static int
qemuPrepareHostUSBDevices(struct qemud_driver *driver,
virDomainDefPtr def)
{
int i, ret = -1;
usbDeviceList *list;
usbDevice *tmp;
virDomainHostdevDefPtr *hostdevs = def->hostdevs;
int nhostdevs = def->nhostdevs;
/* To prevent situation where USB device is assigned to two domains
* we need to keep a list of currently assigned USB devices.
* This is done in several loops which cannot be joined into one big
* loop. See qemuPrepareHostdevPCIDevices()
*/
if (!(list = usbDeviceListNew()))
goto cleanup;
/* Loop 1: build temporary list and validate no usb device
* is already taken
*/
for (i = 0 ; i < nhostdevs ; i++) {
virDomainHostdevDefPtr hostdev = hostdevs[i];
usbDevice *usb = NULL;
if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS)
continue;
if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_USB)
continue;
unsigned vendor = hostdev->source.subsys.u.usb.vendor;
unsigned product = hostdev->source.subsys.u.usb.product;
unsigned bus = hostdev->source.subsys.u.usb.bus;
unsigned device = hostdev->source.subsys.u.usb.device;
if (vendor && bus) {
usb = usbFindDevice(vendor, product, bus, device);
} else if (vendor && !bus) {
usbDeviceList *devs = usbFindDeviceByVendor(vendor, product);
if (!devs)
goto cleanup;
if (usbDeviceListCount(devs) > 1) {
qemuReportError(VIR_ERR_OPERATION_FAILED,
_("multiple USB devices for %x:%x, "
"use <address> to specify one"), vendor, product);
usbDeviceListFree(devs);
goto cleanup;
}
usb = usbDeviceListGet(devs, 0);
usbDeviceListSteal(devs, usb);
usbDeviceListFree(devs);
hostdev->source.subsys.u.usb.bus = usbDeviceGetBus(usb);
hostdev->source.subsys.u.usb.device = usbDeviceGetDevno(usb);
} else if (!vendor && bus) {
usb = usbFindDeviceByBus(bus, device);
}
if (!usb)
goto cleanup;
if (usbDeviceListAdd(list, usb) < 0) {
usbFreeDevice(usb);
goto cleanup;
}
}
/* Mark devices in temporary list as used by @name
* and add them do driver list. However, if something goes
* wrong, perform rollback.
*/
if (qemuPrepareHostdevUSBDevices(driver, def->name, list) < 0)
goto inactivedevs;
/* Loop 2: Temporary list was successfully merged with
* driver list, so steal all items to avoid freeing them
* in cleanup label.
*/
while (usbDeviceListCount(list) > 0) {
tmp = usbDeviceListGet(list, 0);
usbDeviceListSteal(list, tmp);
}
ret = 0;
goto cleanup;
inactivedevs:
/* Steal devices from driver->activeUsbHostdevs.
* We will free them later.
*/
for (i = 0; i < usbDeviceListCount(list); i++) {
tmp = usbDeviceListGet(list, i);
usbDeviceListSteal(driver->activeUsbHostdevs, tmp);
}
cleanup:
usbDeviceListFree(list);
return ret;
}
int qemuPrepareHostdevPCIDevices(struct qemud_driver *driver,
const char *name,
virDomainHostdevDefPtr *hostdevs,
int nhostdevs)
{
pciDeviceList *pcidevs;
int i;
int ret = -1;
if (!(pcidevs = qemuGetPciHostDeviceList(hostdevs, nhostdevs)))
return -1;
/* We have to use 7 loops here. *All* devices must
* be detached before we reset any of them, because
* in some cases you have to reset the whole PCI,
* which impacts all devices on it. Also, all devices
* must be reset before being marked as active.
*/
/* Loop 1: validate that non-managed device isn't in use, eg
* by checking that device is either un-bound, or bound
* to pci-stub.ko
*/
for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
pciDevice *dev = pciDeviceListGet(pcidevs, i);
pciDevice *other;
if (!pciDeviceIsAssignable(dev, !driver->relaxedACS)) {
qemuReportError(VIR_ERR_OPERATION_INVALID,
_("PCI device %s is not assignable"),
pciDeviceGetName(dev));
goto cleanup;
}
/* The device is in use by other active domain if
* the dev is in list driver->activePciHostdevs.
*/
if ((other = pciDeviceListFind(driver->activePciHostdevs, dev))) {
const char *other_name = pciDeviceGetUsedBy(other);
if (other_name)
qemuReportError(VIR_ERR_OPERATION_INVALID,
_("PCI device %s is in use by domain %s"),
pciDeviceGetName(dev), other_name);
else
qemuReportError(VIR_ERR_OPERATION_INVALID,
_("PCI device %s is already in use"),
pciDeviceGetName(dev));
goto cleanup;
}
}
/* Loop 2: detach managed devices */
for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
pciDevice *dev = pciDeviceListGet(pcidevs, i);
if (pciDeviceGetManaged(dev) &&
pciDettachDevice(dev, driver->activePciHostdevs, NULL) < 0)
goto reattachdevs;
}
/* Loop 3: Now that all the PCI hostdevs have been detached, we
* can safely reset them */
for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
pciDevice *dev = pciDeviceListGet(pcidevs, i);
if (pciResetDevice(dev, driver->activePciHostdevs,
driver->inactivePciHostdevs) < 0)
goto reattachdevs;
}
/* Loop 4: Now mark all the devices as active */
for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
pciDevice *dev = pciDeviceListGet(pcidevs, i);
if (pciDeviceListAdd(driver->activePciHostdevs, dev) < 0) {
pciFreeDevice(dev);
goto inactivedevs;
}
}
/* Loop 5: Now remove the devices from inactive list. */
for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
pciDevice *dev = pciDeviceListGet(pcidevs, i);
pciDeviceListDel(driver->inactivePciHostdevs, dev);
}
/* Loop 6: Now set the used_by_domain of the device in
* driver->activePciHostdevs as domain name.
*/
for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
pciDevice *dev, *activeDev;
dev = pciDeviceListGet(pcidevs, i);
activeDev = pciDeviceListFind(driver->activePciHostdevs, dev);
pciDeviceSetUsedBy(activeDev, name);
}
/* Loop 7: Now set the original states for hostdev def */
for (i = 0; i < nhostdevs; i++) {
pciDevice *dev;
pciDevice *pcidev;
virDomainHostdevDefPtr hostdev = hostdevs[i];
if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS)
continue;
if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI)
continue;
dev = pciGetDevice(hostdev->source.subsys.u.pci.domain,
hostdev->source.subsys.u.pci.bus,
hostdev->source.subsys.u.pci.slot,
hostdev->source.subsys.u.pci.function);
/* original states "unbind_from_stub", "remove_slot",
* "reprobe" were already set by pciDettachDevice in
* loop 2.
*/
if ((pcidev = pciDeviceListFind(pcidevs, dev))) {
hostdev->origstates.states.pci.unbind_from_stub =
pciDeviceGetUnbindFromStub(pcidev);
hostdev->origstates.states.pci.remove_slot =
pciDeviceGetRemoveSlot(pcidev);
hostdev->origstates.states.pci.reprobe =
pciDeviceGetReprobe(pcidev);
}
pciFreeDevice(dev);
}
/* Loop 8: Now steal all the devices from pcidevs */
while (pciDeviceListCount(pcidevs) > 0) {
pciDevice *dev = pciDeviceListGet(pcidevs, 0);
pciDeviceListSteal(pcidevs, dev);
}
ret = 0;
goto cleanup;
inactivedevs:
/* Only steal all the devices from driver->activePciHostdevs. We will
* free them in pciDeviceListFree().
*/
while (pciDeviceListCount(pcidevs) > 0) {
pciDevice *dev = pciDeviceListGet(pcidevs, 0);
pciDeviceListSteal(driver->activePciHostdevs, dev);
}
reattachdevs:
for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
pciDevice *dev = pciDeviceListGet(pcidevs, i);
pciReAttachDevice(dev, driver->activePciHostdevs, NULL);
}
cleanup:
pciDeviceListFree(pcidevs);
return ret;
}
qemuAgentPtr
qemuAgentOpen(virDomainObjPtr vm,
virDomainChrSourceDefPtr config,
qemuAgentCallbacksPtr cb)
{
qemuAgentPtr mon;
if (!cb || !cb->eofNotify) {
qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("EOF notify callback must be supplied"));
return NULL;
}
if (VIR_ALLOC(mon) < 0) {
virReportOOMError();
return NULL;
}
if (virMutexInit(&mon->lock) < 0) {
qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("cannot initialize monitor mutex"));
VIR_FREE(mon);
return NULL;
}
if (virCondInit(&mon->notify) < 0) {
qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("cannot initialize monitor condition"));
virMutexDestroy(&mon->lock);
VIR_FREE(mon);
return NULL;
}
mon->fd = -1;
mon->refs = 1;
mon->vm = vm;
mon->cb = cb;
qemuAgentLock(mon);
switch (config->type) {
case VIR_DOMAIN_CHR_TYPE_UNIX:
mon->fd = qemuAgentOpenUnix(config->data.nix.path, vm->pid,
&mon->connectPending);
break;
case VIR_DOMAIN_CHR_TYPE_PTY:
mon->fd = qemuAgentOpenPty(config->data.file.path);
break;
default:
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("unable to handle monitor type: %s"),
virDomainChrTypeToString(config->type));
goto cleanup;
}
if (mon->fd == -1)
goto cleanup;
if ((mon->watch = virEventAddHandle(mon->fd,
VIR_EVENT_HANDLE_HANGUP |
VIR_EVENT_HANDLE_ERROR |
VIR_EVENT_HANDLE_READABLE |
(mon->connectPending ?
VIR_EVENT_HANDLE_WRITABLE :
0),
qemuAgentIO,
mon, qemuAgentUnwatch)) < 0) {
qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("unable to register monitor events"));
goto cleanup;
}
qemuAgentRef(mon);
VIR_DEBUG("New mon %p fd =%d watch=%d", mon, mon->fd, mon->watch);
qemuAgentUnlock(mon);
return mon;
cleanup:
/* We don't want the 'destroy' callback invoked during
* cleanup from construction failure, because that can
* give a double-unref on virDomainObjPtr in the caller,
* so kill the callbacks now.
*/
mon->cb = NULL;
qemuAgentUnlock(mon);
qemuAgentClose(mon);
return NULL;
}
static void
qemuAgentIO(int watch, int fd, int events, void *opaque) {
qemuAgentPtr mon = opaque;
bool error = false;
bool eof = false;
/* lock access to the monitor and protect fd */
qemuAgentLock(mon);
qemuAgentRef(mon);
#if DEBUG_IO
VIR_DEBUG("Agent %p I/O on watch %d fd %d events %d", mon, watch, fd, events);
#endif
if (mon->fd != fd || mon->watch != watch) {
if (events & (VIR_EVENT_HANDLE_HANGUP | VIR_EVENT_HANDLE_ERROR))
eof = true;
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("event from unexpected fd %d!=%d / watch %d!=%d"),
mon->fd, fd, mon->watch, watch);
error = true;
} else if (mon->lastError.code != VIR_ERR_OK) {
if (events & (VIR_EVENT_HANDLE_HANGUP | VIR_EVENT_HANDLE_ERROR))
eof = true;
error = true;
} else {
if (events & VIR_EVENT_HANDLE_WRITABLE) {
if (mon->connectPending) {
if (qemuAgentIOConnect(mon) < 0)
error = true;
} else {
if (qemuAgentIOWrite(mon) < 0)
error = true;
}
events &= ~VIR_EVENT_HANDLE_WRITABLE;
}
if (!error &&
events & VIR_EVENT_HANDLE_READABLE) {
int got = qemuAgentIORead(mon);
events &= ~VIR_EVENT_HANDLE_READABLE;
if (got < 0) {
error = true;
} else if (got == 0) {
eof = true;
} else {
/* Ignore hangup/error events if we read some data, to
* give time for that data to be consumed */
events = 0;
if (qemuAgentIOProcess(mon) < 0)
error = true;
}
}
if (!error &&
events & VIR_EVENT_HANDLE_HANGUP) {
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("End of file from monitor"));
eof = 1;
events &= ~VIR_EVENT_HANDLE_HANGUP;
}
if (!error && !eof &&
events & VIR_EVENT_HANDLE_ERROR) {
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("Invalid file descriptor while waiting for monitor"));
eof = 1;
events &= ~VIR_EVENT_HANDLE_ERROR;
}
if (!error && events) {
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("Unhandled event %d for monitor fd %d"),
events, mon->fd);
error = 1;
}
}
if (error || eof) {
if (mon->lastError.code != VIR_ERR_OK) {
/* Already have an error, so clear any new error */
virResetLastError();
} else {
virErrorPtr err = virGetLastError();
if (!err)
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("Error while processing monitor IO"));
virCopyLastError(&mon->lastError);
virResetLastError();
}
VIR_DEBUG("Error on monitor %s", NULLSTR(mon->lastError.message));
/* If IO process resulted in an error & we have a message,
* then wakeup that waiter */
if (mon->msg && !mon->msg->finished) {
mon->msg->finished = 1;
virCondSignal(&mon->notify);
}
}
qemuAgentUpdateWatch(mon);
/* We have to unlock to avoid deadlock against command thread,
* but is this safe ? I think it is, because the callback
* will try to acquire the virDomainObjPtr mutex next */
if (eof) {
void (*eofNotify)(qemuAgentPtr, virDomainObjPtr)
= mon->cb->eofNotify;
virDomainObjPtr vm = mon->vm;
/* Make sure anyone waiting wakes up now */
virCondSignal(&mon->notify);
if (qemuAgentUnref(mon) > 0)
qemuAgentUnlock(mon);
VIR_DEBUG("Triggering EOF callback");
(eofNotify)(mon, vm);
} else if (error) {
void (*errorNotify)(qemuAgentPtr, virDomainObjPtr)
= mon->cb->errorNotify;
virDomainObjPtr vm = mon->vm;
/* Make sure anyone waiting wakes up now */
virCondSignal(&mon->notify);
if (qemuAgentUnref(mon) > 0)
qemuAgentUnlock(mon);
VIR_DEBUG("Triggering error callback");
(errorNotify)(mon, vm);
} else {
if (qemuAgentUnref(mon) > 0)
qemuAgentUnlock(mon);
}
}
static virJSONValuePtr ATTRIBUTE_SENTINEL
qemuAgentMakeCommand(const char *cmdname,
...)
{
virJSONValuePtr obj;
virJSONValuePtr jargs = NULL;
va_list args;
char *key;
va_start(args, cmdname);
if (!(obj = virJSONValueNewObject()))
goto no_memory;
if (virJSONValueObjectAppendString(obj, "execute", cmdname) < 0)
goto no_memory;
while ((key = va_arg(args, char *)) != NULL) {
int ret;
char type;
if (strlen(key) < 3) {
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("argument key '%s' is too short, missing type prefix"),
key);
goto error;
}
/* Keys look like s:name the first letter is a type code */
type = key[0];
key += 2;
if (!jargs &&
!(jargs = virJSONValueNewObject()))
goto no_memory;
/* This doesn't support maps/arrays. This hasn't
* proved to be a problem..... yet :-) */
switch (type) {
case 's': {
char *val = va_arg(args, char *);
ret = virJSONValueObjectAppendString(jargs, key, val);
} break;
case 'i': {
int val = va_arg(args, int);
ret = virJSONValueObjectAppendNumberInt(jargs, key, val);
} break;
case 'u': {
unsigned int val = va_arg(args, unsigned int);
ret = virJSONValueObjectAppendNumberUint(jargs, key, val);
} break;
case 'I': {
long long val = va_arg(args, long long);
ret = virJSONValueObjectAppendNumberLong(jargs, key, val);
} break;
case 'U': {
/* qemu silently truncates numbers larger than LLONG_MAX,
* so passing the full range of unsigned 64 bit integers
* is not safe here. Pass them as signed 64 bit integers
* instead.
*/
long long val = va_arg(args, long long);
ret = virJSONValueObjectAppendNumberLong(jargs, key, val);
} break;
case 'd': {
double val = va_arg(args, double);
ret = virJSONValueObjectAppendNumberDouble(jargs, key, val);
} break;
case 'b': {
int val = va_arg(args, int);
ret = virJSONValueObjectAppendBoolean(jargs, key, val);
} break;
case 'n': {
ret = virJSONValueObjectAppendNull(jargs, key);
} break;
default:
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("unsupported data type '%c' for arg '%s'"), type, key - 2);
goto error;
}
if (ret < 0)
goto no_memory;
}
if (jargs &&
virJSONValueObjectAppend(obj, "arguments", jargs) < 0)
goto no_memory;
va_end(args);
return obj;
no_memory:
virReportOOMError();
error:
virJSONValueFree(obj);
virJSONValueFree(jargs);
va_end(args);
return NULL;
}
static int
qemuAgentOpenUnix(const char *monitor, pid_t cpid, bool *inProgress)
{
struct sockaddr_un addr;
int monfd;
int timeout = 3; /* In seconds */
int ret, i = 0;
*inProgress = false;
if ((monfd = socket(AF_UNIX, SOCK_STREAM, 0)) < 0) {
virReportSystemError(errno,
"%s", _("failed to create socket"));
return -1;
}
if (virSetNonBlock(monfd) < 0) {
qemuReportError(VIR_ERR_INTERNAL_ERROR,
"%s", _("Unable to put monitor into non-blocking mode"));
goto error;
}
if (virSetCloseExec(monfd) < 0) {
qemuReportError(VIR_ERR_INTERNAL_ERROR,
"%s", _("Unable to set monitor close-on-exec flag"));
goto error;
}
memset(&addr, 0, sizeof(addr));
addr.sun_family = AF_UNIX;
if (virStrcpyStatic(addr.sun_path, monitor) == NULL) {
qemuReportError(VIR_ERR_INTERNAL_ERROR,
_("Agent path %s too big for destination"), monitor);
goto error;
}
do {
ret = connect(monfd, (struct sockaddr *) &addr, sizeof(addr));
if (ret == 0)
break;
if ((errno == ENOENT || errno == ECONNREFUSED) &&
virKillProcess(cpid, 0) == 0) {
/* ENOENT : Socket may not have shown up yet
* ECONNREFUSED : Leftover socket hasn't been removed yet */
continue;
}
if ((errno == EINPROGRESS) ||
(errno == EAGAIN)) {
VIR_DEBUG("Connection attempt continuing in background");
*inProgress = true;
ret = 0;
break;
}
virReportSystemError(errno, "%s",
_("failed to connect to monitor socket"));
goto error;
} while ((++i <= timeout*5) && (usleep(.2 * 1000000) <= 0));
if (ret != 0) {
virReportSystemError(errno, "%s",
_("monitor socket did not show up."));
goto error;
}
return monfd;
error:
VIR_FORCE_CLOSE(monfd);
return -1;
}
int
qemuPrepareHostdevUSBDevices(struct qemud_driver *driver,
const char *name,
virDomainHostdevDefPtr *hostdevs,
int nhostdevs)
{
int ret = -1;
int i;
usbDeviceList *list;
usbDevice *tmp;
/* To prevent situation where USB device is assigned to two domains
* we need to keep a list of currently assigned USB devices.
* This is done in several loops which cannot be joined into one big
* loop. See qemuPrepareHostdevPCIDevices()
*/
if (!(list = usbDeviceListNew()))
goto cleanup;
/* Loop 1: build temporary list and validate no usb device
* is already taken
*/
for (i = 0 ; i < nhostdevs ; i++) {
virDomainHostdevDefPtr hostdev = hostdevs[i];
if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS)
continue;
if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_USB)
continue;
/* Resolve a vendor/product to bus/device */
if (hostdev->source.subsys.u.usb.vendor) {
usbDevice *usb
= usbFindDevice(hostdev->source.subsys.u.usb.vendor,
hostdev->source.subsys.u.usb.product);
if (!usb)
return -1;
hostdev->source.subsys.u.usb.bus = usbDeviceGetBus(usb);
hostdev->source.subsys.u.usb.device = usbDeviceGetDevno(usb);
if ((tmp = usbDeviceListFind(driver->activeUsbHostdevs, usb))) {
const char *other_name = usbDeviceGetUsedBy(tmp);
if (other_name)
qemuReportError(VIR_ERR_OPERATION_INVALID,
_("USB device %s is in use by domain %s"),
usbDeviceGetName(tmp), other_name);
else
qemuReportError(VIR_ERR_OPERATION_INVALID,
_("USB device %s is already in use"),
usbDeviceGetName(tmp));
usbFreeDevice(usb);
goto cleanup;
}
if (usbDeviceListAdd(list, usb) < 0) {
usbFreeDevice(usb);
goto cleanup;
}
}
}
/* Loop 2: Mark devices in temporary list as used by @name
* and add them do driver list. However, if something goes
* wrong, perform rollback.
*/
for (i = 0; i < usbDeviceListCount(list); i++) {
tmp = usbDeviceListGet(list, i);
usbDeviceSetUsedBy(tmp, name);
if (usbDeviceListAdd(driver->activeUsbHostdevs, tmp) < 0) {
usbFreeDevice(tmp);
goto inactivedevs;
}
}
/* Loop 3: Temporary list was successfully merged with
* driver list, so steal all items to avoid freeing them
* in cleanup label.
*/
while (usbDeviceListCount(list) > 0) {
tmp = usbDeviceListGet(list, 0);
usbDeviceListSteal(list, tmp);
}
ret = 0;
goto cleanup;
inactivedevs:
/* Steal devices from driver->activeUsbHostdevs.
* We will free them later.
*/
for (i = 0; i < usbDeviceListCount(list); i++) {
tmp = usbDeviceListGet(list, i);
usbDeviceListSteal(driver->activeUsbHostdevs, tmp);
}
cleanup:
usbDeviceListFree(list);
return ret;
}
int qemuPrepareHostdevPCIDevices(struct qemud_driver *driver,
const char *name,
virDomainHostdevDefPtr *hostdevs,
int nhostdevs)
{
pciDeviceList *pcidevs;
int i;
int ret = -1;
if (!(pcidevs = qemuGetPciHostDeviceList(hostdevs, nhostdevs)))
return -1;
/* We have to use 6 loops here. *All* devices must
* be detached before we reset any of them, because
* in some cases you have to reset the whole PCI,
* which impacts all devices on it. Also, all devices
* must be reset before being marked as active.
*/
/* Loop 1: validate that non-managed device isn't in use, eg
* by checking that device is either un-bound, or bound
* to pci-stub.ko
*/
for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
pciDevice *dev = pciDeviceListGet(pcidevs, i);
pciDevice *other;
if (!pciDeviceIsAssignable(dev, !driver->relaxedACS)) {
qemuReportError(VIR_ERR_OPERATION_INVALID,
_("PCI device %s is not assignable"),
pciDeviceGetName(dev));
goto cleanup;
}
/* The device is in use by other active domain if
* the dev is in list driver->activePciHostdevs.
*/
if ((other = pciDeviceListFind(driver->activePciHostdevs, dev))) {
const char *other_name = pciDeviceGetUsedBy(other);
if (other_name)
qemuReportError(VIR_ERR_OPERATION_INVALID,
_("PCI device %s is in use by domain %s"),
pciDeviceGetName(dev), other_name);
else
qemuReportError(VIR_ERR_OPERATION_INVALID,
_("PCI device %s is already in use"),
pciDeviceGetName(dev));
goto cleanup;
}
}
/* Loop 2: detach managed devices */
for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
pciDevice *dev = pciDeviceListGet(pcidevs, i);
if (pciDeviceGetManaged(dev) &&
pciDettachDevice(dev, driver->activePciHostdevs) < 0)
goto reattachdevs;
}
/* Loop 3: Now that all the PCI hostdevs have been detached, we
* can safely reset them */
for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
pciDevice *dev = pciDeviceListGet(pcidevs, i);
if (pciResetDevice(dev, driver->activePciHostdevs, pcidevs) < 0)
goto reattachdevs;
}
/* Loop 4: Now mark all the devices as active */
for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
pciDevice *dev = pciDeviceListGet(pcidevs, i);
if (pciDeviceListAdd(driver->activePciHostdevs, dev) < 0) {
pciFreeDevice(dev);
goto inactivedevs;
}
}
/* Loop 5: Now set the used_by_domain of the device in
* driver->activePciHostdevs as domain name.
*/
for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
pciDevice *dev, *activeDev;
dev = pciDeviceListGet(pcidevs, i);
activeDev = pciDeviceListFind(driver->activePciHostdevs, dev);
pciDeviceSetUsedBy(activeDev, name);
}
/* Loop 6: Now steal all the devices from pcidevs */
while (pciDeviceListCount(pcidevs) > 0) {
pciDevice *dev = pciDeviceListGet(pcidevs, 0);
pciDeviceListSteal(pcidevs, dev);
}
ret = 0;
goto cleanup;
inactivedevs:
/* Only steal all the devices from driver->activePciHostdevs. We will
* free them in pciDeviceListFree().
*/
while (pciDeviceListCount(pcidevs) > 0) {
pciDevice *dev = pciDeviceListGet(pcidevs, 0);
pciDeviceListSteal(driver->activePciHostdevs, dev);
}
reattachdevs:
for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
pciDevice *dev = pciDeviceListGet(pcidevs, i);
pciReAttachDevice(dev, driver->activePciHostdevs);
}
cleanup:
pciDeviceListFree(pcidevs);
return ret;
}
