[dpdk-dev] Questions Questions about hash function enum

[dpdk-dev] Questions Questions about hash function enum

[oulijun]

Hi Ferruh Yigit and others

    I am learning data structure defined by dpdk framework and I noticed 
a data structure definition below:

enum rte_eth_hash_function {
     RTE_ETH_HASH_FUNCTION_DEFAULT = 0,
     RTE_ETH_HASH_FUNCTION_TOEPLITZ, /**< Toeplitz */
     RTE_ETH_HASH_FUNCTION_SIMPLE_XOR, /**< Simple XOR */
     /**
      * Symmetric Toeplitz: src, dst will be replaced by
      * xor(src, dst). For the case with src/dst only,
      * src or dst address will xor with zero pair.
      */
     RTE_ETH_HASH_FUNCTION_SYMMETRIC_TOEPLITZ,
     RTE_ETH_HASH_FUNCTION_MAX,
};

I have a little difficulty understanding that for the comment:

    /**
      * Symmetric Toeplitz: src, dst will be replaced by
      * xor(src, dst). For the case with src/dst only,
      * src or dst address will xor with zero pair.
      */

if user configure rss func for symmetric toeplitz, Rss type must be 
updated at the same time?

if user configure rss func for xor, rss type must be empty and the queue 
number is zero?

Must it be restricted like this? If so, what is his benefit?

Looking forward to your reply


Thanks

Lijun Ou

Re: [dpdk-dev] Questions Questions about hash function enum

[Ferruh Yigit]

On 6/1/2020 2:31 AM, oulijun wrote:
> Hi Ferruh Yigit and others
> 
>     I am learning data structure defined by dpdk framework and I noticed 
> a data structure definition below:
> 
> enum rte_eth_hash_function {
>      RTE_ETH_HASH_FUNCTION_DEFAULT = 0,
>      RTE_ETH_HASH_FUNCTION_TOEPLITZ, /**< Toeplitz */
>      RTE_ETH_HASH_FUNCTION_SIMPLE_XOR, /**< Simple XOR */
>      /**
>       * Symmetric Toeplitz: src, dst will be replaced by
>       * xor(src, dst). For the case with src/dst only,
>       * src or dst address will xor with zero pair.
>       */
>      RTE_ETH_HASH_FUNCTION_SYMMETRIC_TOEPLITZ,
>      RTE_ETH_HASH_FUNCTION_MAX,
> };
> 
> I have a little difficulty understanding that for the comment:
> 
>     /**
>       * Symmetric Toeplitz: src, dst will be replaced by
>       * xor(src, dst). For the case with src/dst only,
>       * src or dst address will xor with zero pair.
>       */
> 
> if user configure rss func for symmetric toeplitz, Rss type must be 
> updated at the same time?
> 
> if user configure rss func for xor, rss type must be empty and the queue 
> number is zero?
> 
> Must it be restricted like this? If so, what is his benefit?
> 
> Looking forward to your reply

cc'ed Qi & Andrew

As far as I understand that comment is only describing the Symmetric Toeplitz
behavior. I don't see any rss type or queue number limitation.

Re: [dpdk-dev] Questions Questions about hash function enum

[oulijun]

在 2020/6/3 1:13, Ferruh Yigit 写道:
> On 6/1/2020 2:31 AM, oulijun wrote:
>> Hi Ferruh Yigit and others
>>
>>      I am learning data structure defined by dpdk framework and I noticed
>> a data structure definition below:
>>
>> enum rte_eth_hash_function {
>>       RTE_ETH_HASH_FUNCTION_DEFAULT = 0,
>>       RTE_ETH_HASH_FUNCTION_TOEPLITZ, /**< Toeplitz */
>>       RTE_ETH_HASH_FUNCTION_SIMPLE_XOR, /**< Simple XOR */
>>       /**
>>        * Symmetric Toeplitz: src, dst will be replaced by
>>        * xor(src, dst). For the case with src/dst only,
>>        * src or dst address will xor with zero pair.
>>        */
>>       RTE_ETH_HASH_FUNCTION_SYMMETRIC_TOEPLITZ,
>>       RTE_ETH_HASH_FUNCTION_MAX,
>> };
>>
>> I have a little difficulty understanding that for the comment:
>>
>>      /**
>>        * Symmetric Toeplitz: src, dst will be replaced by
>>        * xor(src, dst). For the case with src/dst only,
>>        * src or dst address will xor with zero pair.
>>        */
>>
>> if user configure rss func for symmetric toeplitz, Rss type must be
>> updated at the same time?
>>
>> if user configure rss func for xor, rss type must be empty and the queue
>> number is zero?
>>
>> Must it be restricted like this? If so, what is his benefit?
>>
>> Looking forward to your reply
> 
> cc'ed Qi & Andrew
> 
> As far as I understand that comment is only describing the Symmetric Toeplitz
> behavior. I don't see any rss type or queue number limitation.
> 
Hi Qi & Andrew
    if user configure rss func for symmetric toeplitz, RSS types must be 
updated at the same time?
flow create 0 ingress pattern end actions rss func symmetric_toepilitz 
types tcp end / end

if use the following cmd:
flow create 0 ingress pattern end actions rss func symmetric_toepilitz 
types end / end

it is wrong and the driver will print the error info.
How to understand this limitation?

Thanks
Lijun Ou

> 
> .
>