[EXTERNAL] Re: [v11 0/4] PCI Dev and SG copy support

[fengchengwen]

Hi Gowrishankar,

On 2024/3/12 20:24, Gowrishankar Muthukrishnan wrote:
> Hi Fengchengwen
> 
>>
>> Hi Thomas,
>>
>> On 2024/3/12 17:15, Thomas Monjalon wrote:
>>> 07/03/2024 14:55, Gowrishankar Muthukrishnan:
>>>> Hi Fengchengwen,
>>>>
>>>>>> Waiting for a confirmation that this series is good to go.
>>>>>
>>>>> In the discuss of thread [1], I hope this patchset continue take a
>>>>> step forward (means new version) to support bi-direction test just by
>> modify config.ini file.
>>>>>
>>>>
>>>> This patch set already exposes all configuration via config.ini. I didn't follow
>> what is missing. For bi-direction, we can better continue discussing on that
>> patch.
>>>
>>> Chengwen, please can you confirm whether you require a new version?
>>> Which change exactly is missing?
>>
>> This patchset is OK with one sub-test only tackle one DMA direction.
>>
> Thanks for the confirmation.
> 
>> But there is a later patch [1] which will support multiple DMA directions within
>> one sub-test.
>> it will add a entry "xfer_mode", but I think it complicate the test, I prefer we do
>> more in this patchset to support some like bi-direction just by modify
>> config.ini, some like this:
>>
> I think we should discuss about that in bi-directional patch series. This series is self-contained and there is no need to add bi-directional as part of this series. As far as this patch set is concerned, all the options are exposed via config.ini. Can you comment if there is anything missing, assuming that we are taking bi-directional support as a separate feature addition.

I have identified some improvements to the dma-perf app, and I plan to do it in
24.07, so if you don't mind, I will incorporate your commits (keeping your signed-off-by) and
modify to the one that I described above, and then send to community (also with my improvements
commits).

Thanks

> 
> Thanks,
> Gowrishankar
> 
>> 1. extend lcore_dma:
>>    current lcore_dma is: lcore10 at 0000:00:04.2
>>    extend it support:
>> lcore10 at 0000:00:04.2,dir=m2d,coreid=1,pfid=2,vfid=3,raddr=0xXXXX
>> 2. to fix one entry can't hold too many dma device, support entrys:
>> lcore_dma_1, lcore_dma_2
>>    which value is same with lcore_dma.
>>
>> So for bi-direction, we just define config.ini as:
>> lcore_dma=lcore10 at 0000:00:04.2,dir=m2d,coreid=1,pfid=2,vfid=3,raddr=0x
>> XXXX, lcore10 at 0000:00:04.2,dir=d2m,coreid=1,pfid=2,vfid=3,raddr=0xXXXX
>>
>> [1] https://urldefense.proofpoint.com/v2/url?u=https-
>> 3A__patches.dpdk.org_project_dpdk_patch_20240229141426.4188428-
>> 2D1-2Damitprakashs-
>> 40marvell.com_&d=DwICaQ&c=nKjWec2b6R0mOyPaz7xtfQ&r=EAtr-
>> g7yUFhtOio8r2Rtm13Aqe4WVp_S_gHpcu6KFVo&m=DUaL_AJR1zqM0T2yw3
>> aV44EObB90uqw5weFzSm-
>> w39citSeGozNdEe4kzicss_KG&s=UTAcoZx5DjSJHyzxyLMxXz1bPqfPXQM7feDx
>> ZdC6Jgk&e=
>>
>>>