Re: [dpdk-ci] [DPDKLAB] 25G Broadcom NIC in Arm hardware

[David Liu <dliu@iol.unh.edu>]

[-- Attachment #1: Type: text/plain, Size: 5423 bytes --]

Hi Ajit,

Thank you for helping out.

We have it working on the x86, but not the Arm.


Thanks,
David

On Wed, Jan 13, 2021 at 5:34 PM Ajit Khaparde <ajit.khaparde@broadcom.com>
wrote:

> Hi David,
> I will take a look.
> Do you see similar issues on x86? I am asking because
> even I would start with that to create a baseline and then
> attempt on ARM.
>
> Thanks
> Ajit
>
> On Wed, Jan 13, 2021 at 12:08 PM David Liu <dliu@iol.unh.edu> wrote:
>
>> Hi Ajit,
>>
>> We have a 25G Broadcom NIC that is installed on an Arm machine.
>> And I am running into a problem when I try to run nic_single_core_perf on
>> the NIC.
>> Currently, we are using: Trex v2.86 and dts from
>> http://git.dpdk.org/tools/dts/
>> All the NICs are up and running when testing.
>>
>> The problem is an error when I try to run the test case inside
>> the nic_single_core_perf:
>>
>> TestNicSingleCorePerf: Test running at parameters: framesize: 64,
>>> rxd/txd: 512
>>>                dut.172.18.0.41:
>>> x86_64-native-linuxapp-gcc/app/dpdk-testpmd -l 32,33 -n 4
>>>  --file-prefix=dpdk_15311_20210113190237   -a 0000:93:00.0 -a 0000:93:00.1
>>> -- -i --portmask=0x3 --txd=512 --rxd=512 --nb-cores=1
>>>                dut.172.18.0.41: start
>>>          TestNicSingleCorePerf: Test Case test_perf_nic_single_core
>>> Result ERROR: Traceback (most recent call last):
>>>   File "/opt/dts/framework/test_case.py", line 319, in _execute_test_case
>>>     case_obj()
>>>   File "tests/TestSuite_nic_single_core_perf.py", line 196, in
>>> test_perf_nic_single_core
>>>     self.perf_test(self.nb_ports)
>>>   File "tests/TestSuite_nic_single_core_perf.py", line 270, in perf_test
>>>     _, packets_received =
>>> self.tester.pktgen.measure_throughput(stream_ids=streams,
>>> options=traffic_opt)
>>>   File "/opt/dts/framework/pktgen_base.py", line 245, in
>>> measure_throughput
>>>     self._prepare_transmission(stream_ids=stream_ids)
>>>   File "/opt/dts/framework/pktgen_trex.py", line 778, in
>>> _prepare_transmission
>>>     self._conn.reset(ports=self._ports)
>>>   File
>>> "/opt/trex/v2.86/automation/trex_control_plane/interactive/trex/common/trex_api_annotators.py",
>>> line 51, in wrap2
>>>     ret = f(*args, **kwargs)
>>>   File
>>> "/opt/trex/v2.86/automation/trex_control_plane/interactive/trex/stl/trex_stl_client.py",
>>> line 339, in reset
>>>     self.clear_stats(ports)
>>>   File
>>> "/opt/trex/v2.86/automation/trex_control_plane/interactive/trex/common/trex_api_annotators.py",
>>> line 51, in wrap2
>>>     ret = f(*args, **kwargs)
>>>   File
>>> "/opt/trex/v2.86/automation/trex_control_plane/interactive/trex/stl/trex_stl_client.py",
>>> line 1492, in clear_stats
>>>     self._clear_stats_common(ports, clear_global, clear_xstats)
>>>   File
>>> "/opt/trex/v2.86/automation/trex_control_plane/interactive/trex/common/trex_client.py",
>>> line 2876, in _clear_stats_common
>>>     raise TRexError(rc)
>>> trex.common.trex_exceptions.TRexError: *** [RPC] - Failed to get server
>>> response from tcp://172.18.0.40:4501
>>
>>
>> I also try to run TRex in stateless mode with./t-rex-64 -i --cfg
>> /etc/trex_cfg.yaml and  connect with ./trex-console
>>
>> Then start sending traffic with command
>>
>>> trex>start -f stl/imix.py
>>
>>
>> But this will cause an error to turn off Trex.
>>
>> trex>
>>> 2021-01-13 19:56:21 - [server][warning]  - Server has been shutdown -
>>> cause: 'WATCHDOG: task 'Trex DP core 1' has not responded for more than
>>> 1.00135 seconds - timeout is 1 seconds
>>> *** traceback follows ***
>>> 1       0x561be173cf5a ./_t-rex-64(+0x19af5a) [0x561be173cf5a]
>>> 2       0x7feea3ac0980 /lib/x86_64-linux-gnu/libpthread.so.0(+0x12980)
>>> [0x7feea3ac0980]
>>> 3       0x561be1a05a2a rte_delay_us_block + 106
>>> 4       0x561be16ce874 CCoreEthIF::send_burst(CCorePerPort*, unsigned
>>> short, CVirtualIFPerSideStats*) + 3220
>>> 5       0x561be16abf57 CCoreEthIF::flush_tx_queue() + 31
>>> 6       0x561be171e3d9
>>> CNodeGenerator::handle_maintenance(CFlowGenListPerThread*) + 265
>>> 7       0x561be171f7ec CNodeGenerator::handle_flow_sync(CGenNode*,
>>> CFlowGenListPerThread*, bool&) + 92
>>> 8       0x561be171fc98 CNodeGenerator::handle_slow_messages(unsigned
>>> char, CGenNode*, CFlowGenListPerThread*, bool) + 184
>>> 9       0x561be16cb5f1 int CNodeGenerator::flush_file_realtime<23,
>>> false>(double, double, CFlowGenListPerThread*, double&) + 881
>>> 10      0x561be1905212 TrexStatelessDpCore::start_scheduler() + 226
>>> 11      0x561be1883ae9 TrexDpCore::start() + 89
>>> 12      0x561be1714113
>>> CFlowGenListPerThread::start(std::__cxx11::basic_string<char,
>>> std::char_traits<char>, std::allocator<char> >&, CPreviewMode&) + 115
>>> 13      0x561be16af8dd CGlobalTRex::run_in_core(unsigned char) + 487
>>> 14      0x561be16d11ad ./_t-rex-64(+0x12f1ad) [0x561be16d11ad]
>>> 15      0x561be1a1dfaa eal_thread_loop + 426
>>> 16      0x7feea3ab56db /lib/x86_64-linux-gnu/libpthread.so.0(+0x76db)
>>> [0x7feea3ab56db]
>>> 17      0x7feea2a8571f clone + 63
>>>
>>> *** addr2line information follows ***
>>> ??:0
>>> ??:0
>>> ??:0
>>> ??:0
>>> ??:0
>>> ??:0
>>> ??:0
>>> ??:0
>>> ??:0
>>> ??:0
>>> ??:0
>>> ??:0
>>> ??:0
>>> ??:0
>>> ??:0
>>> ??:0
>>> ??:0
>>> '
>>
>>
>> I believe if everything is working, this should not turn off trex but
>> please correct me if I am wrong.
>> I wonder if you have any suggestion of how I can approach this issue?
>>
>> Best,
>> David
>>
>>
>>

[-- Attachment #2: Type: text/html, Size: 16261 bytes --]