Re: [patch net-next 10/10] selftests: netdevsim: Extend devlink trap test to include flow action cookie

[Jiri Pirko <jiri@resnulli.us>]

Tue, Feb 25, 2020 at 06:57:21PM CET, kuba@kernel.org wrote:
>On Tue, 25 Feb 2020 08:46:03 +0100 Jiri Pirko wrote:
>> >>  		devlink_trap_metadata_test $trap_name "input_port"
>> >>  		check_err $? "Input port not reported as metadata of trap $trap_name"
>> >> +		if [ $trap_name == "ingress_flow_action_drop" ] ||
>> >> +		   [ $trap_name == "egress_flow_action_drop" ]; then
>> >> +			devlink_trap_metadata_test $trap_name "flow_action_cookie"
>> >> +			check_err $? "Flow action cookie not reported as metadata of trap $trap_name"
>> >> +		fi
>> >>  	done
>> >>  
>> >>  	log_test "Trap metadata"  
>> >
>> >Oh, this doesn't seem to check the contents of the trap at all, does it?  
>> 
>> No. This is not the test for the actual trapped packets. It is a test to
>> list devlink traps and supported metadata.
>> 
>> The packet trapping is done using dropmonitor which is currently
>> not implemented in selftests, even for the existing traps. Not even for
>> mlxsw. There is a plan to introduce these tests in the future, Ido is
>> working on a tool to catch those packets to pcap using dropmon. I think
>> he plans to send it to dropmon git soon. Then we can implement the
>> selftests using it.
>
>The extra 100 lines of code in netdevsim which is not used by selftests
>does make me a little sad.. but okay, looking forward to fuller tests.
>Those tests better make use of the variable cookie size, 'cause
>otherwise we could have just stored the cookie on a u64 and avoided the
>custom read/write functions all together ;)

Will do.