Re: [Qemu-devel] [PATCH 4/5] vfio: Move container list to DMAContext

[Alexey Kardashevskiy <aik@ozlabs.ru>]

On 04/26/2013 06:52 PM, Paolo Bonzini wrote:
> Il 26/04/2013 10:46, Alexey Kardashevskiy ha scritto:
>> On 04/26/2013 06:44 PM, Paolo Bonzini wrote:
>>> Il 25/04/2013 08:36, David Gibson ha scritto:
>>>> On Wed, Apr 24, 2013 at 06:33:33PM +0200, Paolo Bonzini wrote:
>>>>> Il 24/04/2013 17:12, Alex Williamson ha scritto:
>>>>>>>> At the moment, vfio maintains a global list of containers
>>>>>>>> that are assumed to be more or less interchangeable, since
>>>>>>>> they are all set up with a MemoryListener to have all of
>>>>>>>> system memory mapped.  However, that only makes sense if
>>>>>>>> all the containers are used on devices which really do
>>>>>>>> expect a dma address space identical to system memory.
>>>>>>>>
>>>>>>>> This patch moves towards that by making the list of
>>>>>>>> containers per DMAContext (which corresponds to a dma
>>>>>>>> address space) instead of global.
>>>>>> This seems like an unnecessary intrusion into common code.  Why
>>>>>> not create a vfio specific list of dma objects, each with a
>>>>>> list of containers?  Thanks,
>>>>>
>>>>> Yeah, I suggest that this is re-evaluated on top of the iommu
>>>>> patches.
>>>>
>>>>> You can find them at git://github.com/bonzini/qemu.git, branch
>>>>> iommu. It seems to work with pseries, at least my guest crashes
>>>>> at the same place with and without.  USB works, and so do VGA and
>>>>> spapr-vscsi.
>>>>
>>>> Ok, I'll have a look when I get a chance.  Any guesses as to when
>>>> they might reach mainline?
>>>
>>> If I get your Tested-by, early in 1.6.
>>
>> Emulated PCI works on ppc64/spapr so far.
>
> What about VIO?  Doesn't it go through the IOMMU as well?  I'm sure you
> can test it more than I did (it did break in Avi's original patches, so
> it must not be that bad...).


I run QEMU with
-net nic,model=ibmveth -net user,hostfwd=tcp::5000-:22
and run in the guest: "dhclient ; wget google.com" which worked fine. When 
it damaged, it does not go further than dhclient :)



-- 
Alexey