[libvirt] [PATCH 07/13] qemu: Add hotpluging support for PCI devices on S390 guests

[Yi Min Zhao]

在 2018/5/29 下午5:04, Cornelia Huck 写道:
> On Mon, 28 May 2018 16:26:57 +0800
> Xiao Feng Ren <renxiaof at linux.vnet.ibm.com> wrote:
>
>> On 5/25/2018 6:22 PM, Bjoern Walk wrote:
>>> Cornelia Huck <cohuck at redhat.com> [2018-05-24, 06:25PM +0200]:
>>>> On Thu, 24 May 2018 14:24:32 +0200
>>>> Xiao Feng Ren <renxiaof at linux.vnet.ibm.com> wrote:
>>>>   
>>>>> From: Yi Min Zhao <zyimin at linux.ibm.com>
>>>>>
>>>>> This commit adds hotplug support for PCI devices on S390 guests.
>>>>> There's no need to implement hot unplug for zPCI as QEMU implements
>>>>> an unplug callback which will unplug both PCI and zPCI device in a
>>>>> cascaded way.
>>>>> Currently, the following PCI devices are supported:
>>>>>     virtio-blk-pci
>>>>>     virtio-net-pci
>>>>>     virtio-rng-pci
>>>>>     virtio-input-host-pci
>>>>>     virtio-keyboard-pci
>>>>>     virtio-mouse-pci
>>>>>     virtio-tablet-pci
>>>>>     vfio-pci
>>>>>     Shmem device
>>>>>     SCSIVhost device
>>>> Hm, how did you arrive at this list? Is it 'anything that uses msi-x'?
>>> I guess it's just any device that libvirt actually supports hotplug for,
>>> with a few exceptions (cf. patch 3 and 6). Not familiar wuth msi-x.
>> The list should be the devices that support MSI-X and have realized the
>> hotplug in qemu,
>> we support the hotplug for them in the libvirt.  So the list will be
>> updated to:
>>
>>     virtio-blk-pci
>>     virtio-net-pci
>>     virtio-input-host-pci
>>     virtio-keyboard-pci
>>     virtio-mouse-pci
>>     virtio-tablet-pci
>>     vfio-pci
>>     SCSIVhost device
> Ok, that makes sense.
>
> I also checked that libvirt only allows setting the vectors property
> for virtio-serial and shmem (vectors == 0 turns msi-x off), so that
> should be fine.

Thanks for your check!
>
>> As the model of Shmem is not support in qemu, the rng device doesn't
>> support MSI-X.  So remove them.
> So, should qemu add support for msi-x with virtio-rng-pci in the
> future, we'd want to update libvirt as well, correct?
This is a little bit hard to answer. Our patch adds the code calling the 
function which is to
attach zpci address in rng attachment function. As the logic in libvirt, 
zpci could be generated
for rng device. But it actually can't be supported in s390 qemu binary. 
If it is being plugged,
the error should be reported.

Thus, if we remove the attachement code for zpci in 
qemuDomainAttachRNGDevice(),
we should update libvirt when rng supports msix in qemu. If we keep it 
here, then update for libvirt
is not needed.
>
>>
>> In addition, the bug found need to be fixed for the hotplug of tablet,
>> in the virDomainDeviceIsUSB()
>> if ((t == VIR_DOMAIN_DEVICE_DISK &&
>>            dev->data.disk->bus == VIR_DOMAIN_DISK_BUS_USB) ||
>>           (t == VIR_DOMAIN_DEVICE_INPUT &&
>>            dev->data.input->type == VIR_DOMAIN_INPUT_BUS_USB) ||
>> ---------->dev->data.input->bus == VIR_DOMAIN_INPUT_BUS_USB
>>           (t == VIR_DOMAIN_DEVICE_HOSTDEV &&
>> I will send the fix if this bug is agreed.
>>
>