[libvirt PATCH v2] qemu: ramfb video device doesn't support PCI address

Laine Stump laine at redhat.com
Thu Jun 25 16:43:35 UTC 2020


On 6/25/20 12:30 PM, Jonathon Jongsma wrote:
> On Thu, 2020-06-25 at 12:20 -0400, Laine Stump wrote:
>> On 6/25/20 10:34 AM, Jonathon Jongsma wrote:
>>> Although a ramfb video device is not a PCI device, we don't
>>> currently
>>> report an error for ramfb device definitions containing a PCI
>>> address.
>>> However, a guest configured with such a device will fail to start:
>>>
>>>       # virsh start test1
>>>       error: Failed to start domain test1
>>>       error: internal error: qemu unexpectedly closed the monitor:
>>> 2020-06-16T05:23:02.759221Z qemu-kvm: -device
>>> ramfb,id=video0,bus=pcie.0,addr=0x1: Device 'ramfb' can't go on
>>> PCIE bus
>>>
>>> A better approach is to reject any device definitions that contain
>>> PCI
>>> addresses.  While this is a change in behavior, any existing
>>> configurations were non-functional.
>>>
>>> https://bugzilla.redhat.com/show_bug.cgi?id=1847259
>>>
>>> Signed-off-by: Jonathon Jongsma <jjongsma at redhat.com>
>>> ---
>>> changes in v2:
>>>    - move validation to qemu-specific validation function as
>>> suggested by Laine
>>>
>>>    src/qemu/qemu_validate.c | 8 ++++++++
>>>    tests/qemuxml2argvtest.c | 1 +
>>>    2 files changed, 9 insertions(+)
>>>
>>> diff --git a/src/qemu/qemu_validate.c b/src/qemu/qemu_validate.c
>>> index 5082a29dc7..b13c03759e 100644
>>> --- a/src/qemu/qemu_validate.c
>>> +++ b/src/qemu/qemu_validate.c
>>> @@ -1925,6 +1925,14 @@ qemuValidateDomainDeviceDefVideo(const
>>> virDomainVideoDef *video,
>>>        if (qemuValidateDomainVirtioOptions(video->virtio, qemuCaps)
>>> < 0)
>>>            return -1;
>>>    
>>> +    if (video->type == VIR_DOMAIN_VIDEO_TYPE_RAMFB &&
>>> +        video->info.type != VIR_DOMAIN_DEVICE_ADDRESS_TYPE_NONE) {
>>> +        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
>>> +                       _("'address' is not supported for 'ramfb'
>>> video devices"));
>>> +        return -1;
>>> +    }
>>> +
>>> +
>>>        return 0;
>>>    }
>>>    
>>> diff --git a/tests/qemuxml2argvtest.c b/tests/qemuxml2argvtest.c
>>> index 1195f9c982..f2522fa530 100644
>>> --- a/tests/qemuxml2argvtest.c
>>> +++ b/tests/qemuxml2argvtest.c
>>> @@ -2276,6 +2276,7 @@ mymain(void)
>>>                QEMU_CAPS_VIRTIO_GPU_MAX_OUTPUTS);
>>>        DO_TEST_CAPS_LATEST("video-bochs-display-device");
>>>        DO_TEST_CAPS_LATEST("video-ramfb-display-device");
>>> +    DO_TEST_CAPS_LATEST_PARSE_ERROR("video-ramfb-display-device-
>>> pci-address");
>> Did you forget to git-add the test case data?
> OK, well that points out an interesting property of the
> DO_TEST_CAPS_LATEST_PARSE_ERROR() macro. It passed the 'make check'
> because it failed to parse the xml file as expected. But the reason it
> failed was obviously not the reason I expected it to fail... I want it
> to fail due to the specifying a PCI address for the ramfb device. But
> it actually fails because the xml file didn't exist... :/


That gave me the best laugh I'd had all day!! :-)


I just realized you don't have push privileges on gitlab. If you re-send 
with test case data, I'll push it.


(I guess we should also fix the test harness to "fail to fail" if the 
test case data is missing, but that's a separate issue)




More information about the libvir-list mailing list