[dpdk-dev] [PATCH v4 0/6] net/mlx5: add Rx buffer split support

Ferruh Yigit ferruh.yigit at intel.com
Thu Oct 29 14:09:45 CET 2020


On 10/27/2020 7:05 PM, Slava Ovsiienko wrote:
>> -----Original Message-----
>> From: Ferruh Yigit <ferruh.yigit at intel.com>
>> Sent: Tuesday, October 27, 2020 13:05
>> To: Slava Ovsiienko <viacheslavo at nvidia.com>; Raslan Darawsheh
>> <rasland at nvidia.com>; dev at dpdk.org
>> Cc: NBU-Contact-Thomas Monjalon <thomas at monjalon.net>; Matan Azrad
>> <matan at nvidia.com>; Alexander Kozyrev <akozyrev at nvidia.com>; Ori Kam
>> <orika at nvidia.com>
>> Subject: Re: [dpdk-dev] [PATCH v4 0/6] net/mlx5: add Rx buffer split support
>>
>> On 10/26/2020 5:38 PM, Slava Ovsiienko wrote:
>>> Hi,  Ferruh
>>>
>>> PSB
>>>> -----Original Message-----
>>>> From: Ferruh Yigit <ferruh.yigit at intel.com>
>>>> Sent: Monday, October 26, 2020 19:04
>>>> To: Raslan Darawsheh <rasland at nvidia.com>; Slava Ovsiienko
>>>> <viacheslavo at nvidia.com>; dev at dpdk.org
>>>> Cc: NBU-Contact-Thomas Monjalon <thomas at monjalon.net>; Matan Azrad
>>>> <matan at nvidia.com>; Alexander Kozyrev <akozyrev at nvidia.com>; Ori Kam
>>>> <orika at nvidia.com>
>>>> Subject: Re: [dpdk-dev] [PATCH v4 0/6] net/mlx5: add Rx buffer split
>>>> support
>>>>
>>>> On 10/26/2020 3:25 PM, Raslan Darawsheh wrote:
>>>>> Hi,
>>>>>
>>>>>> -----Original Message-----
>>>>>> From: Viacheslav Ovsiienko <viacheslavo at nvidia.com>
>>>>>> Sent: Monday, October 26, 2020 1:55 PM
>>>>>> To: dev at dpdk.org
>>>>>> Cc: NBU-Contact-Thomas Monjalon <thomas at monjalon.net>; Matan
>> Azrad
>>>>>> <matan at nvidia.com>; Alexander Kozyrev <akozyrev at nvidia.com>;
>> Raslan
>>>>>> Darawsheh <rasland at nvidia.com>; Ori Kam <orika at nvidia.com>
>>>>>> Subject: [PATCH v4 0/6] net/mlx5: add Rx buffer split support
>>>>>>
>>>>>> This patch adds to PMD the functionality for the receiving buffer
>>>>>> split feasture [1]
>>>>>>
>>>>>> [1]
>>>>>> https://nam11.safelinks.protection.outlook.com/?url=http%3A%2F%2Fpa
>>>>>> tc
>>>>>> h
>>>>
>> es.dpdk.org%2Fpatch%2F81154%2F&data=02%7C01%7Crasland%40nvid
>>>>>>
>>>>
>> ia.com%7Ccf4913c6b58346b50b1b08d879a60608%7C43083d15727340c1b7db
>>>> 3
>>>>>>
>>>>
>> 9efd9ccc17a%7C0%7C0%7C637393101256743078&sdata=fyiL3PS8r8wv8u
>>>>>> pyOYUtITkVqId9DZsF9LvSJQL9fdM%3D&reserved=0
>>>>>>
>>>>>> Signed-off-by: Viacheslav Ovsiienko <viacheslavo at nvidia.com>
>>>>>>
>>>>>> ---
>>>>>> v1:
>>>>>> https://nam11.safelinks.protection.outlook.com/?url=http%3A%2F%2Fpa
>>>>>> tc
>>>>>> h
>>>>
>> es.dpdk.org%2Fpatch%2F81808%2F&data=02%7C01%7Crasland%40nvid
>>>>>>
>>>>
>> ia.com%7Ccf4913c6b58346b50b1b08d879a60608%7C43083d15727340c1b7db
>>>> 3
>>>>>>
>>>>
>> 9efd9ccc17a%7C0%7C0%7C637393101256743078&sdata=NPBFlGmVN6bi
>>>>>> GUpzHC%2FrOVmdMoK2fkYRC0%2FDB%2BNlNno%3D&reserved=0
>>>>>>
>>>>>> v2:
>>>>>> https://nam11.safelinks.protection.outlook.com/?url=http%3A%2F%2Fpa
>>>>>> tc
>>>>>> h
>>>>
>> es.dpdk.org%2Fpatch%2F81923%2F&data=02%7C01%7Crasland%40nvid
>>>>>>
>>>>
>> ia.com%7Ccf4913c6b58346b50b1b08d879a60608%7C43083d15727340c1b7db
>>>> 3
>>>>>>
>>>>
>> 9efd9ccc17a%7C0%7C0%7C637393101256743078&sdata=YwYjMz3jrSYU6
>>>>>> RBgwl0DmQfmjwwymNJTFjMdx0rsm2U%3D&reserved=0
>>>>>>        - typos
>>>>>>        - documentation is updated
>>>>>>
>>>>>> v3:
>>>>>> https://nam11.safelinks.protection.outlook.com/?url=http%3A%2F%2Fpa
>>>>>> tc
>>>>>> h
>>>>
>> es.dpdk.org%2Fpatch%2F82177%2F&data=02%7C01%7Crasland%40nvid
>>>>>>
>>>>
>> ia.com%7Ccf4913c6b58346b50b1b08d879a60608%7C43083d15727340c1b7db
>>>> 3
>>>>>>
>>>>
>> 9efd9ccc17a%7C0%7C0%7C637393101256743078&sdata=HVvLbWS0sJxu
>>>>>> v%2Bc%2BKIMqllBq3edC4v0GD%2BtrwS7%2FsRo%3D&reserved=0
>>>>>>        - extra parameter checks in PMD rx_queue_setup removed
>>>>>>        - minor optimizations in PMD
>>>>>>
>>>>>> v4: - rebasing
>>>>>>
>>>>>> Viacheslav Ovsiienko (6):
>>>>>>      net/mlx5: add extended Rx queue setup routine
>>>>>>      net/mlx5: configure Rx queue to support split
>>>>>>      net/mlx5: register multiple pool for Rx queue
>>>>>>      net/mlx5: update Rx datapath to support split
>>>>>>      net/mlx5: report Rx segmentation capabilities
>>>>>>      doc: add buffer split feature limitation to mlx5 guide
>>>>>>
>>>>>>     doc/guides/nics/mlx5.rst        |   6 +-
>>>>>>     drivers/net/mlx5/mlx5.h         |   3 +
>>>>>>     drivers/net/mlx5/mlx5_ethdev.c  |   4 ++
>>>>>>     drivers/net/mlx5/mlx5_mr.c      |   3 +
>>>>>>     drivers/net/mlx5/mlx5_rxq.c     | 136
>>>>>> +++++++++++++++++++++++++++++++++++-----
>>>>>>     drivers/net/mlx5/mlx5_rxtx.c    |   3 +-
>>>>>>     drivers/net/mlx5/mlx5_rxtx.h    |  13 +++-
>>>>>>     drivers/net/mlx5/mlx5_trigger.c |  20 +++---
>>>>>>     8 files changed, 160 insertions(+), 28 deletions(-)
>>>>>>
>>>>>> --
>>>>>> 1.8.3.1
>>>>>
>>>>> Series applied to next-net-mlx,
>>>>>
>>>>
>>>> The feature was references with different name in each commit, I
>>>> tried to unify it as "Rx buffer split" in next-net.
>>>> Can you please double check the updated commit log/titles?
>>>
>>>>> 	doc: add Rx buffer split limitation to mlx5 guide
>>>>> 	net/mlx5: report Rx buffer split capabilities
>>> OK about above.
>>>
>>>>> 	net/mlx5: support Rx buffer split
>>> It would be better: "net/mlx5: support Rx buffer split on datapath
>>>
>>
>> Isn't the supporting the "Rx buffer split" mean supporting it on the datapath,
>> where else it can be supported, the "on datapath" looks redundant to me.
> 
> Options for possible "support Buffer Split" meaning:
> - generic PMD configuration
> - queue configuration
> - reporting caps
> - datapath
> 
> The series is split for commits those updating the very specific parts in PMD.
> We may drop this specifics but we would lose the series split meaning.
> Sure, the entire series is about "support Rx buffer split", but each commit has
> its own clarification in the headline.
> 
>>
>>>>> 	net/mlx5: register multiple pool for Rx queue
>>> OK
>>>
>>>>> 	net/mlx5: configure Rx buffer split
>>> It would be better: "net/mlx5: configure Rx queue for buffer split"
>>>
>>
>> Like above, isn't the configure "Rx buffer split" mean configuring Rx queue for
>> it, "Rx queue" looks redundant to me.
> It just emphasizes - "the queue object is configured in this specific commit",
> it would be easier to find this point and understand what it is in the long git log.
> Hence, in my opinion, "queue" is some kind of extra clue, we should not drop it.
> 
>> For both above, if you have strong opinion to update them, I can. But I prefer
>> shorter versions.
>>
> 
>>>>> 	net/mlx5: receive Rx buffer split description
>>> IMO, it would be better: "net/mlx5: handle Rx buffer split description"
>>> or
>>> "net/mlx5: support Rx buffer split description"
>>>
>>
>> OK to use "net/mlx5: support Rx buffer split description"
>>
> Please, see dpdk-next-net-mlx - Raslan updated the subtree, addressing
> the hotfix and yours and mine comments.
> 

What to see in the sub-tree?
Making changes is easy, the essence is discussion and reaching into a consensus, 
which is what I am trying to do, without a consensus what is the point of 
updating it in the mlx sub-tree?


More information about the dev mailing list