[libvirt] [PATCH 2/3] qemu: improve device flag parsing

[Wen Congyang]

At 01/19/2011 09:54 AM, Eric Blake Write:
> On 01/18/2011 06:22 PM, Wen Congyang wrote:
>> At 01/18/2011 11:16 PM, Eric Blake Write:
>>> On 01/17/2011 10:39 PM, Wen Congyang wrote:
>>>>> +    cmd = virCommandNewArgList(qemu,
>>>>> +                               "-device", "pci-assign,?",
>>>>> +                               NULL);
>>>>
>>>> The qemu that I used does not support '-device pci-assign,?'...
>>>> So, I can not start the guest with this patch.
>>>>
>>>> The qemu is cloned from here:
>>>> http://git.qemu.org/git/qemu.git
>>>
>>> What do the following commands produce (assuming that both qemu and
>>> qemu-kvm on your PATH are the binaries that you are testing with)?
>>
>> After builing qemu, I only have the qemy-system-x86_64 command
> 
> That explains it.  git.qemu.org/git/qemu.git is the upstream qemu
> repository, and lacks kvm integration.  As such, it lacks -device
> pci-assign.
> 
> To build qemu-kvm, you also need to track
> git://git.kernel.org/pub/scm/virt/kvm/qemu-kvm.git (which is heavily
> based on the former qemu.git).  Without kvm support, qemu will still be
> able to emulate x86_64 platforms, but it's noticeably slower.

The qemu in git.qemu.org/git/qemu.git has supported kvm.

> 
>>
>> # qemu-system-x86_64 --version
>> QEMU emulator version 0.13.50, Copyright (c) 2003-2008 Fabrice Bellard
> 
> Good - my reproduction of your issue (and my patch) was correct.
> 
>> # qemu-system-x86_64 -device ? -device pci-assign,? -device virtio-blk-pci,?
>> name "virtio-9p-pci", bus PCI
> ...
>> name "isa-debugcon", bus ISA
>> virtio-blk-pci.class=hex32
>> virtio-blk-pci.drive=drive
>> virtio-blk-pci.logical_block_size=uint16
>> virtio-blk-pci.physical_block_size=uint16
>> virtio-blk-pci.min_io_size=uint16
>> virtio-blk-pci.opt_io_size=uint32
>> virtio-blk-pci.bootindex=int32
>> virtio-blk-pci.discard_granularity=uint32
>> virtio-blk-pci.vectors=uint32
>> virtio-blk-pci.indirect_desc=on/off
>> virtio-blk-pci.scsi=on/off
> 
> And this, as desired, is a successful feature probe even though it lacks
> pci-assign.* information.
> 
> Sorry for getting your name wrong in my commit message.
It does not matter.
>