Re: [PATCH 2/6] net/hns3: fix inconsistent enabled RSS behavior

[Ferruh Yigit <ferruh.yigit@xilinx.com>]

On 3/9/2022 9:55 AM, Ori Kam wrote:
> Hi Lihusiong,
> 
>> -----Original Message-----
>> From: lihuisong (C) <lihuisong@huawei.com>
>> Sent: Wednesday, March 9, 2022 10:03 AM
>> Subject: Re: [PATCH 2/6] net/hns3: fix inconsistent enabled RSS behavior
>>
>>
>> 在 2022/3/3 10:47, lihuisong (C) 写道:
>>> Hi, Ori,
>>>
>>> 在 2022/3/2 22:07, Ori Kam 写道:
>>>> Hi Lihuisong,
>>>>
>>>>> -----Original Message-----
>>>>> From: lihuisong (C) <lihuisong@huawei.com>
>>>>> Sent: Wednesday, March 2, 2022 4:10 AM
>>>>> Subject: Re: [PATCH 2/6] net/hns3: fix inconsistent enabled RSS
>>>>> behavior
>>>>>
>>>>>
>>>>> 在 2022/3/1 0:42, Ferruh Yigit 写道:
>>>>>> On 2/28/2022 3:21 AM, Min Hu (Connor) wrote:
>>>>>>> From: Huisong Li <lihuisong@huawei.com>
>>>>>>>
>>>>>>> RSS will not be enabled if the RTE_ETH_MQ_RX_RSS_FLAG isn't be set in
>>>>>>> dev_configure phase. However, if this flag isn't set, RSS can be
>>>>>>> enabled
>>>>>>> through the ethdev ops and rte_flow API. This behavior is contrary to
>>>>>>> each
>>>>>>> other.
>>>>>>>
>>>>>>> Fixes: c37ca66f2b27 ("net/hns3: support RSS")
>>>>>>> Cc: stable@dpdk.org
>>>>>>>
>>>>>>> Signed-off-by: Huisong Li <lihuisong@huawei.com>
>>>>>>
>>>>>> Hi Huisong, Connor,
>>>>>>
>>>>>> Let's get a little more feedback for this patch, cc'ed more people.
>>>>>>
>>>>>>
>>>>>> To enable RSS, multi queue mode should be set to
>>>>>> 'RTE_ETH_MQ_RX_RSS_FLAG'.
>>>>>>
>>>>>> But I wonder if it is required to configure RSS via flow API,
>>>>> I do not know the original purpose of adding the RSS configuration in
>>>>> flow API.
>>>>>
>>>> The purpose is simple, this allow to create RSS per rule and not a
>>>> global one.
>>>> For example create RSS that sends TCP to some queues while othe RSS
>>>> will send
>>>> UDP traffic to different queues.
>>> I'm a little confused now. The "per rule" also seems to be a global
>>> configuration.
>>> Example:
>>>   - start PMD with 0,1,2,3
>>>   - create TCP packets to 2,3 queues. At this moment, only 2,3 queues
>>> can be received for other types of packets.
>>> Because this rule is implemented by modifying the entry of the
>>> redirection table which is global for this device.
>> Hi, Ori and Stephen.
>> Can you help me clear up the confusion above? If some NICs behave like
>> this, what should we do about it?
> 
> I'm not sure I understand the issue, maybe it is releated to some HW/PMD limitation.
> In your example non TCP traffic will be routed to one of the 4 queues (0,1,2,3),
> While TCP traffic will only be routed to queues 2,3.
> 
> Now I can add new rule that matches on UDP packet and RSS to queue 0 and 3 in this case:
> TCP packets will be routed to queues 0,3.
> UDP packets will be routed to queues 2,3.
> All the rest of the traffic will be routed to queues 0,1,2,3
> 

Hi Ori,

How RETA is managed for per flow RSS configuration? And is there a limit 
on how many per flow RSS rule can be created?

> And just to be clear if now I add a rule to match all packets in higher priority,
> with RSS to queues 1,2. Then all traffic will be routed to queues 1,2.
> 
> At least this is what is expected, from API point of view.
> 
> Best,
> Ori
> 
>>>>
>>>>> However, as far as I know, the hash algorithm can be configured via
>>>>> this
>>>>> API,
>>>>>
>>>>> but not via ethdev ops API.
>>>>>
>>>>>> and if other PMDs check this configuration for flow API?
>>>>> Some PMDs already have similar check in RSS releated ops or rte_flow
>>>>> API.
>>>>>
>>>>> For example, hinic, axbge, bnxt, cnxk, otx2, and ena.
>>>>>
>>>>>> .
>>>>   From my view point those are two different settings.
>>>> The RTE_ETH_MQ_RX_RSS_FLAG is global per port while
>>>> rte_flow is per rule.
>>>>
>>>> I think, that if a PMD needs this flag, in order to enable it also
>>>> for rte_flow then
>>>> it should be documented in the release note of the PMD.
>>>> It is a valid use case that the default traffic will not have RSS and
>>>> only rules created by
>>>> rte_flow will have the RSS, for matching traffc.
>>>>
>>>> Best,
>>>> Ori