Re: [PATCH v7 1/9] ethdev: overwrite some comment related to RSS

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

On 10/28/2023 3:41 AM, lihuisong (C) wrote:
> 
> 在 2023/10/28 9:46, Jie Hai 写道:
>> 1. overwrite the comments of fields of 'rte_eth_rss_conf'.
>> 2. Add comments for RTE_ETH_HASH_FUNCTION_DEFAULT.
>>
>> Signed-off-by: Jie Hai <haijie1@huawei.com>
>> ---
>>   lib/ethdev/rte_ethdev.h | 33 ++++++++++++++++++---------------
>>   lib/ethdev/rte_flow.h   |  1 +
>>   2 files changed, 19 insertions(+), 15 deletions(-)
>>
>> diff --git a/lib/ethdev/rte_ethdev.h b/lib/ethdev/rte_ethdev.h
>> index 2fd3cd808dbf..37fd5afef48a 100644
>> --- a/lib/ethdev/rte_ethdev.h
>> +++ b/lib/ethdev/rte_ethdev.h
>> @@ -448,24 +448,27 @@ struct rte_vlan_filter_conf {
>>   /**
>>    * A structure used to configure the Receive Side Scaling (RSS) feature
>>    * of an Ethernet port.
>> - * If not NULL, the *rss_key* pointer of the *rss_conf* structure points
>> - * to an array holding the RSS key to use for hashing specific header
>> - * fields of received packets. The length of this array should be
>> indicated
>> - * by *rss_key_len* below. Otherwise, a default random hash key is
>> used by
>> - * the device driver.
>> - *
>> - * The *rss_key_len* field of the *rss_conf* structure indicates the
>> length
>> - * in bytes of the array pointed by *rss_key*. To be compatible, this
>> length
>> - * will be checked in i40e only. Others assume 40 bytes to be used as
>> before.
>> - *
>> - * The *rss_hf* field of the *rss_conf* structure indicates the
>> different
>> - * types of IPv4/IPv6 packets to which the RSS hashing must be applied.
>> - * Supplying an *rss_hf* equal to zero disables the RSS feature.
>>    */
>>   struct rte_eth_rss_conf {
>> -    uint8_t *rss_key;    /**< If not NULL, 40-byte hash key. */
>> +    /**
>> +     * In rte_eth_dev_rss_hash_conf_get(), the *rss_key_len* should be
>> +     * greater than or equal to the hash_key_size which get from
>>
> Is this added the new comment?
> Suggest that the "rss_key_len" field should be equal to the
> hash_key_size from dev_info_get().
> Because many PMDs, like, hns3, i40e, ice and mlx5, check it in driver
> based on the condition that the rss_key_len field must be equal to the
> hash_key_size when the rss_key is not NULL.
> IMO, it is better that this check should be added in ethdev layer.
>

+1 to add check in ethdev layer


>> +     * rte_eth_dev_info_get() API. And the *rss_key* should contain
>> at least
>> +     * *rss_key_len* bytes. If not meet these requirements, the query
>> result
> 
> here "rss_key_len" should be "hash_key_size", right?
> 
>> +     * is unreliable even the operation returns success.
> s/even/even if
>> +     *
>> +     * In rte_eth_dev_rss_hash_update() or rte_eth_dev_configure(), the
>> +     * *rss_key_len* indicates the length of the *rss_key* in bytes of
>> +     * the array pointed by *rss_key*. Drivers are free to ignore the
>> +     * *rss_key_len* and assume key length is 40 bytes.
>>
> please modify the comment: "Drivers are free to ignore the *rss_key_len*
> and assume key length is 40 bytes. "
> Actually, most of PMDs, like, hns3, i40e, mlx, do not configure RSS hash
> key if rss_key is NULL and treat it as not updating the hash key.
>

This is not for the case where rss_key is NULL, most driver use
hardcoded 40 bytes as rss key len, so they don't check the rss_key_len,
only some drivers use not 40 bytes length, that is why above comment is
correct I think.


>> +     */
>> +    uint8_t *rss_key;
>>       uint8_t rss_key_len; /**< hash key length in bytes. */
>> -    uint64_t rss_hf;     /**< Hash functions to apply - see below. */
>> +    /**
>> +     * Indicates the type of packets or the specific part of packets to
>> +     * which RSS hashing is to be applied.
>> +     */
>> +    uint64_t rss_hf;
>>   };
>>     /*
>> diff --git a/lib/ethdev/rte_flow.h b/lib/ethdev/rte_flow.h
>> index edefa34c10da..25f1dffd1f30 100644
>> --- a/lib/ethdev/rte_flow.h
>> +++ b/lib/ethdev/rte_flow.h
>> @@ -3226,6 +3226,7 @@ struct rte_flow_query_count {
>>    * Hash function types.
>>    */
>>   enum rte_eth_hash_function {
>> +    /** DEFAULT means driver decides which hash algorithm to pick. */
>>       RTE_ETH_HASH_FUNCTION_DEFAULT = 0,
>>       RTE_ETH_HASH_FUNCTION_TOEPLITZ, /**< Toeplitz */
>>       RTE_ETH_HASH_FUNCTION_SIMPLE_XOR, /**< Simple XOR */