Re: [RFC 2/2] vhost: IFC VF vdpa layer

From: Jason Wang <jasowang@redhat.com>
To: "Zhu, Lingshan" <lingshan.zhu@intel.com>,
	Zhu Lingshan <lingshan.zhu@linux.intel.com>,
	mst@redhat.com, alex.williamson@redhat.com
Cc: linux-kernel@vger.kernel.org,
	virtualization@lists.linux-foundation.org, kvm@vger.kernel.org,
	netdev@vger.kernel.org, dan.daly@intel.com,
	cunming.liang@intel.com, tiwei.bie@intel.com,
	jason.zeng@intel.com, zhiyuan.lv@intel.com
Subject: Re: [RFC 2/2] vhost: IFC VF vdpa layer
Date: Wed, 23 Oct 2019 14:39:15 +0800	[thread overview]
Message-ID: <02d44f0a-687f-ed87-518b-7a4d3e83c5d3@redhat.com> (raw)
In-Reply-To: <6588d9f4-f357-ec78-16a4-ccaf0e3768e7@intel.com>

On 2019/10/23 下午2:19, Zhu, Lingshan wrote:
>
> On 10/22/2019 9:05 PM, Jason Wang wrote:
>>
>> On 2019/10/22 下午2:53, Zhu Lingshan wrote:
>>>
>>> On 10/21/2019 6:19 PM, Jason Wang wrote:
>>>>
>>>> On 2019/10/21 下午5:53, Zhu, Lingshan wrote:
>>>>>
>>>>> On 10/16/2019 6:19 PM, Jason Wang wrote:
>>>>>>
>>>>>> On 2019/10/16 上午9:30, Zhu Lingshan wrote:
>>>>>>> This commit introduced IFC VF operations for vdpa, which complys to
>>>>>>> vhost_mdev interfaces, handles IFC VF initialization,
>>>>>>> configuration and removal.
>>>>>>>
>>>>>>> Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
>>>>>>> ---
>>
>>
>> [...]
>>
>>
>>>>
>>>>
>>>>>
>>>>>>
>>>>>>
>>>>>>> +}
>>>>>>> +
>>>>>>> +static int ifcvf_mdev_set_features(struct mdev_device *mdev, 
>>>>>>> u64 features)
>>>>>>> +{
>>>>>>> +    struct ifcvf_adapter *adapter = mdev_get_drvdata(mdev);
>>>>>>> +    struct ifcvf_hw *vf = IFC_PRIVATE_TO_VF(adapter);
>>>>>>> +
>>>>>>> +    vf->req_features = features;
>>>>>>> +
>>>>>>> +    return 0;
>>>>>>> +}
>>>>>>> +
>>>>>>> +static u64 ifcvf_mdev_get_vq_state(struct mdev_device *mdev, 
>>>>>>> u16 qid)
>>>>>>> +{
>>>>>>> +    struct ifcvf_adapter *adapter = mdev_get_drvdata(mdev);
>>>>>>> +    struct ifcvf_hw *vf = IFC_PRIVATE_TO_VF(adapter);
>>>>>>> +
>>>>>>> +    return vf->vring[qid].last_avail_idx;
>>>>>>
>>>>>>
>>>>>> Does this really work? I'd expect it should be fetched from hw 
>>>>>> since it's an internal state.
>>>>> for now, it's working, we intend to support LM in next version 
>>>>> drivers.
>>>>
>>>>
>>>> I'm not sure I understand here, I don't see any synchronization 
>>>> between the hardware and last_avail_idx, so last_avail_idx should 
>>>> not change.
>>>>
>>>> Btw, what did "LM" mean :) ?
>>>
>>> I can add bar IO operations here, LM = live migration, sorry for the 
>>> abbreviation.
>>
>>
>> Just make sure I understand here, I believe you mean reading 
>> last_avail_idx through IO bar here?
>>
>> Thanks
>
> Hi Jason,
>
> Yes, I mean last_avail_idx. is that correct?
>
> THanks

Yes.

Thanks

>
>>
>>