Re: [libvirt] [PATCH v3 22/24] qemu: Use modern vcpu hotplug approach if possible

To allow unplugging the vcpus, hotplugging of vcpus on platforms which require to plug multiple logical vcpus at once or plugging them in an arbitrary order it's necessary to use the new device_add interface for vcpu hotplug. This patch adds support for the device_add interface using the old setvcpus API by implementing an algorithm to select the appropriate entities to plug in. --- src/qemu/qemu_driver.c | 156 +++++++++++++++++++++++++++++++++++++++++++------ 1 file changed, 139 insertions(+), 17 deletions(-)

cleanup: + virJSONValueFree(vcpuprops); return ret; } @@ -4771,6 +4792,95 @@ qemuDomainSetVcpusMax(virQEMUDriverPtr driver, } +/** + * qemuDomainSelectHotplugVcpuEntities: + * + * @def: domain definition + * @nvcpus: target vcpu count + * @cpumap: vcpu entity IDs filled on success

+ * + * Tries to find which vcpu entities need to be enabled or disabled to reach + * @nvcpus. This function works in order of the legacy hotplug but is able to + * skip over entries that are added out of order. + */ +static virBitmapPtr +qemuDomainSelectHotplugVcpuEntities(virDomainDefPtr def, + unsigned int nvcpus) +{ + virBitmapPtr ret = NULL; + virDomainVcpuDefPtr vcpu; + qemuDomainVcpuPrivatePtr vcpupriv; + unsigned int maxvcpus = virDomainDefGetVcpusMax(def); + unsigned int curvcpus = virDomainDefGetVcpus(def); + ssize_t i; +