Re: [libvirt] [PATCH 03/14] qemu: Introduce qemuDomainChrDefPostParse()
On Thu, 2017-11-16 at 11:30 +0100, Pavel Hrdina wrote:
> @@ -4096,13 +4109,6 @@
qemuDomainDeviceDefPostParse(virDomainDeviceDefPtr dev,
> }
> }
>
> - /* set the default console type for S390 arches */
> - if (dev->type == VIR_DOMAIN_DEVICE_CHR &&
> - dev->data.chr->deviceType == VIR_DOMAIN_CHR_DEVICE_TYPE_CONSOLE
&&
> - dev->data.chr->targetType == VIR_DOMAIN_CHR_CONSOLE_TARGET_TYPE_NONE
&&
> - ARCH_IS_S390(def->os.arch))
> - dev->data.chr->targetType =
VIR_DOMAIN_CHR_CONSOLE_TARGET_TYPE_VIRTIO;
> -
You've missed the following condition that clears auto generated unix
socket path, it is also for char devices.
Right. I was focusing on the target type, but the stuff you mention
should definitely go in qemuDomainChrDefPostParse() too.
> @@ -4154,6 +4160,11 @@
qemuDomainDeviceDefPostParse(virDomainDeviceDefPtr dev,
> qemuDomainShmemDefPostParse(dev->data.shmem) < 0)
> goto cleanup;
>
> + if (dev->type == VIR_DOMAIN_DEVICE_CHR &&
> + qemuDomainChrDefPostParse(dev->data.chr, def) < 0) {
> + goto cleanup;
> + }
> +
Is there any specific reason why did you move it to a different place?
All the ad-hoc code is before the sub-functions, and it looks nicer
to have the sub-functions all together. There should be no functional
impact with moving it.
--
Andrea Bolognani / Red Hat / Virtualization