[dpdk-dev] [RFC PATCH] vhost: remove lockless enqueue to virt ring

[Xie, Huawei]

On 11/6/2015 9:43 AM, Tetsuya Mukawa wrote:
> On 2015/11/05 19:21, Xie, Huawei wrote:
>> If no objections, would remove the internal cmpset enqueue, should get
>> some improved performance.
>>
>> No idea why commit message is lost:
>> DPDK doesn't support receiving/transmitting packets from/to the same
>> port/queue.
>> Vhost pmd wraps vhost device as normal dpdk port. dpdkvhost based
>> applications/switch
>>  also treate virtio port and physical port both as ports, and have
>> their  own
>> lock implmentation when to enqueue packets to the same port. This patch
>> removes
>> the internal lockless vhost enqueue implmentation.
>> On 11/5/2015 5:38 PM, Huawei Xie wrote:
>>> Signed-off-by: Huawei Xie <huawei.xie at intel.com>
> Hi Xie,
>
> I think it's a good optimization.
> But I  guess the patch will changes behavior of vhost library API.
> For example, if someone has already implemented an application based on
> the assumption that vhost library has internal lock, I doubt their
> application will face an issue with this patch.
Yes, i understand. If we have an improper implementation, we fix it as
early as possible. I prefer to remove the internal lock currently. The
last thing i want to do is maintain two APIs.
Would follow the ABI process.
>
> If it's correct, we might need to follow below steps.
> http://dpdk.org/doc/guides/contributing/versioning.html
>
> Thanks,
> Tetsuya
>