On 07/18/2011 06:27 PM, Jiri Denemark wrote:
Call qemu monitor command directly within a special job that is only allowed during outgoing migration. --- src/qemu/qemu_domain.c | 1 + src/qemu/qemu_domain.h | 3 +-- src/qemu/qemu_driver.c | 23 +++++++++++++++-------- src/qemu/qemu_migration.c | 21 +++++---------------- src/qemu/qemu_process.c | 1 + 5 files changed, 23 insertions(+), 26 deletions(-)
- VIR_DEBUG("Requesting migration speed change to %luMbs", bandwidth); - priv->job.signalsData.migrateBandwidth = bandwidth; - priv->job.signals |= QEMU_JOB_SIGNAL_MIGRATE_SPEED; - ret = 0; + VIR_DEBUG("Setting migration bandwidth to %luMbs", bandwidth); + ignore_value(qemuDomainObjEnterMonitor(driver, vm));
Might be nice to add a comment about why qemuDomainObjEnterMonitor won't fail here (if I understand correctly, it is because we only get this far if priv->job.active represents a migration job, but qemuDomainObjEnterMonitorInternal can only fail if priv->job.active is QEMU_JOB_NONE), but this isn't the first time this file uses ignore_value() without comments. ACK. -- Eric Blake eblake@redhat.com +1-801-349-2682 Libvirt virtualization library http://libvirt.org