Re: [Qemu-devel] [PATCH v2 0/6] vhost-user: Specify and implement device IOTLB support

[Maxime Coquelin <maxime.coquelin@redhat.com>]

On 06/01/2017 09:04 AM, Jason Wang wrote:
> 
> 
> On 2017年05月31日 23:32, Maxime Coquelin wrote:
>>>>> [0]:https://gitlab.com/mcoquelin/dpdk-next-virtio/commits/vhost_iotlb_proto_v2 
>>>>>
>>>>> [1]:https://lists.gnu.org/archive/html/qemu-devel/2016-04/msg00095.html 
>>>>>
>>>> Overall, this looks good to me. I do think patch 3 isn't a good idea
>>>> though, if slave wants something let it request it.
>>>>
>>>> Need to find out why does vhost in kernel want the used ring iotlb at
>>>> start time - especially considering we aren't even guaranteed one entry
>>>> covers the whole ring, and invalidates should affect all addresses at
>>>> least in theory.
>>>>
>>>>
>>>
>>> The reason is probably we want to verify whether or not we could 
>>> correctly access used ring in vhost_vq_init_access(). It was there 
>>> since vhost_net is introduced. We can think to remove this limitation 
>>> maybe.
>>
>> Even if we remove the limitation on Kernel side, we will still have to
>> keep this workaround for compatibility with older kernels. Having done 
>> the test, I can confirm it is currently necessary.
>>
>> Thanks,
>> Maxime 
> 
> Right, it was probably too late for the change.

So, I wonder if we could keep patch 3, that just extends this
workaround for vhost-user. Else, we would need to do the ring-related 
translations in the data path, which will be costly.

Michael, what do you think?

Thanks,
Maxime