ENRT/ConfigMixins/DevNfcRxFlowHashConfigMixin.py: add support for missing keys

[jtluka]

Description

Resolves the issue we've seen in downstream (job id 12174708) when the ethtool's output for nfc-rx-flow-hash reported an unsupported key.

The fix is to add support for the missing keys.

Tests

Job id 12184448 Job id v2 12186672

Tests newly added settings:

• "L2DA": "m",
• "L3 proto": "t",
• "VLAN tag": "v",

Reviews

@olichtne

[jtluka]

Downstream job contains an issue, this will need additional analysis and fix.

[jtluka]

So it seems that the ethtool output completely changed the format.

On the left the new format breaking LNST parsing, on the right the old format:

 '_res': {'passed': True,                                                                      | '_res': {'passed': True,                                                                      
          'res_data': {'stderr': '',                                                           |          'res_data': {'stderr': '',                                            
                       'stdout': 'UDP over IPV4 flows use these fields for '                   |                       'stdout': 'UDP over IPV4 flows use these fields for '                   
                                 'computing Hash flow key:\n'                                  |                                 'computing Hash flow key:\n'                                  
                                 'L2DA\n'                                                      |                                 'IP SA\n'                                                     
                                 'L3 proto\n'                                                  |                                 'IP DA\n'                                                     
                                 '\n'},                                                        |                                 'L4 bytes 0 & 1 [TCP/UDP src port]\n'                         
          'type': 'result'},                                                                   |                                 'L4 bytes 2 & 3 [TCP/UDP dst port]\n'                         
 '_what': 'ethtool -n ens4f0np0 rx-flow-hash udp4'}                                            |                                 '\n'},                                                        
{'_desc': None,                                                                                |          'type': 'result'},                                                                   

[jtluka]

So it seems that the ethtool output completely changed the format.

On the left the new format breaking LNST parsing, on the right the old format:

 '_res': {'passed': True,                                                                      | '_res': {'passed': True,                                                                      
          'res_data': {'stderr': '',                                                           |          'res_data': {'stderr': '',                                            
                       'stdout': 'UDP over IPV4 flows use these fields for '                   |                       'stdout': 'UDP over IPV4 flows use these fields for '                   
                                 'computing Hash flow key:\n'                                  |                                 'computing Hash flow key:\n'                                  
                                 'L2DA\n'                                                      |                                 'IP SA\n'                                                     
                                 'L3 proto\n'                                                  |                                 'IP DA\n'                                                     
                                 '\n'},                                                        |                                 'L4 bytes 0 & 1 [TCP/UDP src port]\n'                         
          'type': 'result'},                                                                   |                                 'L4 bytes 2 & 3 [TCP/UDP dst port]\n'                         
 '_what': 'ethtool -n ens4f0np0 rx-flow-hash udp4'}                                            |                                 '\n'},                                                        
{'_desc': None,                                                                                |          'type': 'result'},                                                                   

I've checked the ethtool source code and the L2DA and L3 proto are supported. It could be that the NICs for some reason have different values configured.

We should add support for all possible settings.

[jtluka]

So I checked the Intel ice and Nvidia mlx5_core NIC whether it is possible to set any of the "m", "t", or "v" and none of these can be set with ethtool reporting back with "Invalid argument error.

With that I somehow doubt that this MR is actually needed.

[jtluka]

So I checked the Intel ice and Nvidia mlx5_core NIC whether it is possible to set any of the "m", "t", or "v" and none of these can be set with ethtool reporting back with "Invalid argument error.

With that I somehow doubt that this MR is actually needed.

bnxt_en should support "v" and "m". I will do a test run on that NIC to check the MR correctness.

[jtluka]

So I checked the Intel ice and Nvidia mlx5_core NIC whether it is possible to set any of the "m", "t", or "v" and none of these can be set with ethtool reporting back with "Invalid argument error.
With that I somehow doubt that this MR is actually needed.

bnxt_en should support "v" and "m". I will do a test run on that NIC to check the MR correctness.

testing in J:12248420

[jtluka]

So I checked the Intel ice and Nvidia mlx5_core NIC whether it is possible to set any of the "m", "t", or "v" and none of these can be set with ethtool reporting back with "Invalid argument error.
With that I somehow doubt that this MR is actually needed.

bnxt_en should support "v" and "m". I will do a test run on that NIC to check the MR correctness.

testing in J:12248420

I had to update the test configurations however none of m/v/t cannot be configured on the bnxt_en NICs we have. Since these values are not commonly used, I will remove support for these options as they're not really needed and only include the patch to not raise error but instead just log an error message.

[jtluka]

I will completely drop this MR. The patches were kind of workaround for a broken ethtool kernel code, so there's no real need of it ...