[dpdk-dev] [PATCH] net/i40e: fix incorrect hash look up table

Jeff Guo jia.guo at intel.com
Sat Jul 18 06:07:08 CEST 2020


hi, shougang

On 7/15/2020 5:07 PM, Chen, BoX C wrote:
> Tested-by: zhang,xi <xix.zhang at intel.com>
>
> Regards,
> Chen Bo
>
>> -----Original Message-----
>> From: dev <dev-bounces at dpdk.org> On Behalf Of Shougang Wang
>> Sent: July 15, 2020 14:35
>> To: dev at dpdk.org
>> Cc: Xing, Beilei <beilei.xing at intel.com>; Guo, Jia <jia.guo at intel.com>; Wang,
>> ShougangX <shougangx.wang at intel.com>; stable at dpdk.org
>> Subject: [dpdk-dev] [PATCH] net/i40e: fix incorrect hash look up table
>>
>> The hash look up table(LUT) will not be initializing when starting testpmd with
>> --disable-rss. So that some queues in LUT will be invalid if rx queues is less
>> than last time. When enable RSS by creating RSS rule, some packets will not
>> be into the valid queues.


What does "So that some queues in LUT will be invalid if rx queues is 
less than last time" means, Could you explain more clear?


>> This patch fixes this issue by initializing the LUT when creating an RSS rule.
>>
>> Fixes: feaae285b342 ("net/i40e: support hash configuration in RSS flow")
>> Cc: stable at dpdk.org
>>
>> Signed-off-by: Shougang Wang <shougangx.wang at intel.com>
>> ---
>>   drivers/net/i40e/i40e_ethdev.c | 52
>> ++++++++++++++++++++++++++++++++++
>>   1 file changed, 52 insertions(+)
>>
>> diff --git a/drivers/net/i40e/i40e_ethdev.c b/drivers/net/i40e/i40e_ethdev.c
>> index 393b5320f..2a92bd8ef 100644
>> --- a/drivers/net/i40e/i40e_ethdev.c
>> +++ b/drivers/net/i40e/i40e_ethdev.c
>> @@ -13070,6 +13070,49 @@ i40e_rss_conf_init(struct
>> i40e_rte_flow_rss_conf *out,
>>   return 0;
>>   }
>>
>> +static int
>> +i40e_rss_init_lut(struct i40e_pf *pf)


I saw that this function is almost the same functional as 
i40e_rss_config_queue_region, could you check if it could be reused and 
merged one.


>> +{
>> +struct i40e_hw *hw = I40E_PF_TO_HW(pf);
>> +uint32_t lut = 0;
>> +uint16_t j, num;
>> +uint32_t i;
>> +
>> +/*
>> + * If both VMDQ and RSS enabled, not all of PF queues are configured.
>> + * It's necessary to calculate the actual PF queues that are configured.
>> + */
>> +if (pf->dev_data->dev_conf.rxmode.mq_mode &
>> ETH_MQ_RX_VMDQ_FLAG)
>> +num = i40e_pf_calc_configured_queues_num(pf);
>> +else
>> +num = pf->dev_data->nb_rx_queues;
>> +
>> +num = RTE_MIN(num, I40E_MAX_Q_PER_TC);
>> +PMD_INIT_LOG(INFO, "Max of contiguous %u PF queues are
>> configured",
>> +     num);
>> +
>> +if (num == 0) {
>> +PMD_INIT_LOG(ERR,
>> +"No PF queues are configured to enable RSS for
>> port %u",
>> +pf->dev_data->port_id);
>> +return -ENOTSUP;
>> +}
>> +
>> +if (pf->adapter->rss_reta_updated == 0) {
>> +for (i = 0, j = 0; i < hw->func_caps.rss_table_size; i++, j++) {
>> +if (j == num)
>> +j = 0;
>> +lut = (lut << 8) | (j & ((0x1 <<
>> +hw->func_caps.rss_table_entry_width) - 1));
>> +if ((i & 3) == 3)
>> +I40E_WRITE_REG(hw, I40E_PFQF_HLUT(i >>
>> 2),
>> +       rte_bswap32(lut));
>> +}
>> +}
>> +
>> +return 0;
>> +}
>> +
>>   /* Write HENA register to enable hash */  static int  i40e_rss_hash_set(struct
>> i40e_pf *pf, struct i40e_rte_flow_rss_conf *rss_conf) @@ -13318,12
>> +13361,21 @@ static int  i40e_rss_enable_hash(struct i40e_pf *pf,
>>   struct i40e_rte_flow_rss_conf *conf)
>>   {
>> +enum rte_eth_rx_mq_mode mq_mode =
>> +pf->dev_data->dev_conf.rxmode.mq_mode;
>>   struct i40e_rte_flow_rss_conf *rss_info = &pf->rss_info;
>>   struct i40e_rte_flow_rss_conf rss_conf;
>>
>>   if (!(conf->conf.types & pf->adapter->flow_types_mask))
>>   return -ENOTSUP;
>>
>> +/*
>> + * If the RSS is disabled before this, the LUT is uninitialized.
>> + * So it is necessary to initialize it here.
>> + */


Please don't use an empty /* line, use /* Comment....


>> +if (!(mq_mode & ETH_MQ_RX_RSS_FLAG) && !pf-
>>> rss_info.conf.queue_num)
>> +if (i40e_rss_init_lut(pf))
>> +return -ENOTSUP;


How can it know it is not support in function's caller,  better to use 
variable to restore the calling return and return the variable.


>> +
>>   memset(&rss_conf, 0, sizeof(rss_conf));
>>   rte_memcpy(&rss_conf, conf, sizeof(rss_conf));
>>
>> --
>> 2.17.1


More information about the dev mailing list