[dpdk-dev] [PATCH] app/testpmd: fix error detection in MTU command

[Ferruh Yigit]

On 5/19/2020 5:40 PM, Shy Shyman wrote:
> Hi 
> Thank you for the feedback.
> While I'm ok with the fix you suggested, the warning was placed in order to let the user understand that his/her MTU settings may not impact DPDK. After all in this case(no JUMBO) we are not doing anything besides changing the dev->data->mtu.

Does anything needs to changed in the no jumbo case? I think no, but if there is
why no change it instead of warning message. I think warning can be confusing here.

> 
> I think there is some confusion with MTU in DPDK and I'd like to take the opportunity the raise some questions related to MTU:
> 
> 1. only the test-pmd application attempts to correlate MTU to other variables like max_ptk_len and JUMBO offload. should rte_ethdev  do that?
> 2. even in test-pmd case only RX settings are adjusted. Maybe TX should be regarded as well?
> 3. in general what should be the relation between MTU,max_pkt_let and JUMBO offload for both RX and TX?
> 4. currently the MTU is set upon initialization to 1500 by default. The MTU on the device itself could be different but DPDK doesn't check for that. Maybe it should be updated based on the device settings? On a similar note, let's say the user changed MTU in test-pmd. What is the desired behavior when existing the application? Today the MTU will remain changed even outside DPDK.
> 
> Thoughts?

+1 on the confusion. I don't have answers to all above question, but it would be
great if you tackle this clarification task.

> 
> 
> Shy
> 
> -----Original Message-----
> From: Ferruh Yigit <ferruh.yigit at intel.com> 
> Sent: Tuesday, May 19, 2020 5:58 PM
> To: Shy Shyman <shys at mellanox.com>; dev at dpdk.org
> Cc: Wenzhuo Lu <wenzhuo.lu at intel.com>; Beilei Xing <beilei.xing at intel.com>; Bernard Iremonger <bernard.iremonger at intel.com>; xavier.huwei at huawei.com
> Subject: Re: [dpdk-dev] [PATCH] app/testpmd: fix error detection in MTU command
> 
> On 5/18/2020 10:27 AM, Shy Shyman wrote:
>> MTU is used in testpmd to set the maximum payload size for packets.
>> According to testpmd, the setting influnce RX only.
>> In rte_ethdev there's no relation between MTU setting and JUMBO 
>> offload or rx_max_pkt_len.
>>
>> The previous fix in patch referenced below was meant to update the 
>> correlated variables of max_pkt_len and JUMBO offload, but by doing so 
>> it assumes that MTU setting can only exist when JUMBO offload 
>> supported in the device.
> 
> +1 and this is wrong, as far as I understand this is not the intention 
> +of the
> previous fix.
> 
> 
>> For example fail-safe device does supports set MTU and doesn't support 
>> JUMBO offload, and in this case, though set MTU succeed an error 
>> mesage still printed  since the JUMBO packet offload is disabled.
>>
>> The fix separates the two conditions to make sure the error triggers 
>> only in case the set_mtu action actually failed.
>> A warning message is provided in this special case to alert the user.
> 
> Not sure if this warning is required at all.
> As you said above intention is to based on MTU value to correlate testpmd
> (application) max_pkt_len and JUMBO offload values.
> 
> What about following change:
> 
> .....
> 	diag = rte_eth_dev_set_mtu(port_id, mtu);
> 	if (diag) {
> 		printf("Set MTU failed. diag=%d\n", diag);
> 		return;
> 	}
> 
> 	if (dev_info.rx_offload_capa & DEV_RX_OFFLOAD_JUMBO_FRAME) {
> 		....
> 		....
> 		....
> 	}
> }
> 
> 
>>
>> Fixes: 150c9ac2df13 ("app/testpmd: update Rx offload after setting 
>> MTU")
>> Cc: xavier.huwei at huawei.com
>>
>> Signed-off-by: Shy Shyman <shys at mellanox.com>
>> ---
>>  app/test-pmd/config.c | 12 ++++++------
>>  1 file changed, 6 insertions(+), 6 deletions(-)
>>
>> diff --git a/app/test-pmd/config.c b/app/test-pmd/config.c index 
>> 5381207cc..73b53c50b 100644
>> --- a/app/test-pmd/config.c
>> +++ b/app/test-pmd/config.c
>> @@ -1277,8 +1277,9 @@ port_mtu_set(portid_t port_id, uint16_t mtu)
>>  		return;
>>  	}
>>  	diag = rte_eth_dev_set_mtu(port_id, mtu);
>> -	if (diag == 0 &&
>> -	    dev_info.rx_offload_capa & DEV_RX_OFFLOAD_JUMBO_FRAME) {
>> +	if (diag)
>> +		printf("Set MTU failed. diag=%d\n", diag);
>> +	else if (dev_info.rx_offload_capa & DEV_RX_OFFLOAD_JUMBO_FRAME) {
>>  		/*
>>  		 * Ether overhead in driver is equal to the difference of
>>  		 * max_rx_pktlen and max_mtu in rte_eth_dev_info when the @@ 
>> -1293,10 +1294,9 @@ port_mtu_set(portid_t port_id, uint16_t mtu)
>>  		} else
>>  			rte_port->dev_conf.rxmode.offloads &=
>>  						~DEV_RX_OFFLOAD_JUMBO_FRAME;
>> -
>> -		return;
>> -	}
>> -	printf("Set MTU failed. diag=%d\n", diag);
>> +	} else
>> +		printf("WARNING: MTU was set while jumbo frame offload is"
>> +			" not supported by the device\n");
>>  }
>>  
>>  /* Generic flow management functions. */
>>
>