Skip to content

example-tcpip.md

Latest commit

 

History

History
187 lines (141 loc) · 4.67 KB

example-tcpip.md

File metadata and controls

187 lines (141 loc) · 4.67 KB

ExampleTCPIP

ExampleTCPIP is a hardware stress example for:

  • ServerSocket (TCP server)
  • Socket (TCP client from board to host)
  • DatagramSocket (UDP)
  • Server connection manager

It includes command/control, payload integrity checks (checksum), burst/saturation traffic, forced disconnect/reconnect and UDP probe/ack.

Control command IDs used by the host script:

  • CMD_RESET=1
  • CMD_PING=2
  • CMD_GET_STATS=3
  • CMD_FORCE_DISCONNECT=4
  • CMD_BURST_SERVER=5
  • CMD_BURST_CLIENT=6
  • CMD_FORCE_CLIENT_RECONNECT=7
  • CMD_GET_HEALTH=8 (paged telemetry)
  • CMD_RESET_HEALTH=9

1. Build

Build with Ethernet enabled and EXAMPLE_TCPIP defined.

Example (board + KSZ8041):

./hyper build tcpip --preset board-debug-eth-ksz8041 --extra-cxx-flags "-DTCPIP_TEST_HOST_IP=192.168.1.9"

Example (nucleo + LAN8742):

./hyper build tcpip --preset nucleo-debug-eth --extra-cxx-flags "-DTCPIP_TEST_HOST_IP=192.168.1.9"

Notes:

  • TCPIP_TEST_HOST_IP must be the IPv4 of your laptop on the same Ethernet segment.
  • Defaults (if not overridden in compile flags):
    • TCPIP_TEST_BOARD_IP="192.168.1.7"
    • TCP server port: 40000
    • TCP client local/remote ports: 40001 / 40002
    • UDP local/remote ports: 40003 / 40004

2. Flash and run

Flash the latest build with:

./hyper flash

Then power-cycle or reset the board if needed.

One-shot automation (build + flash + ping + tests):

./tools/run_example_tcpip_nucleo.sh \
  --iface en6 \
  --board-ip 192.168.1.7 \
  --base-runs 1 \
  --aggr-runs 0

3. Run stress tests from laptop

./tools/run_example_tcpip_stress.sh --board-ip 192.168.1.7 --host-bind 192.168.1.9

Useful options:

  • --host-bind 0.0.0.0
  • --tcp-server-port 40000
  • --tcp-client-port 40002
  • --udp-local-port 40003
  • --udp-remote-port 40004
  • --good-payloads 1200
  • --bad-payloads 200
  • --min-payload-rx-ratio 0.90
  • --min-bad-detect-ratio 0.80
  • --payload-interval-us 800 (set 0 for max blast / likely RX overrun testing)
  • --server-burst 800
  • --client-burst 800
  • --min-server-burst-ratio 0.95
  • --min-client-burst-ratio 0.95
  • --udp-count 300
  • --strict-client-stream (make tcp_client_stream a hard fail instead of warning)
  • --health-pages 6
  • --reset-health
  • --health-at-end
  • --no-health-on-fail

4. What the script validates

  • TCP command/response path (PING)
  • TCP payload integrity under load (good + bad checksum packets)
  • Forced disconnect and reconnect
  • TCP server burst stream reception
  • UDP probe/ack response ratio and board-reported counters
  • Board TCP client stream reception on host-side sink

5. Telemetry Pages (CMD_GET_HEALTH)

CMD_GET_HEALTH returns three values per page:

  • page 0: uptime_ms, loop_iterations, tcp_commands_rx
  • page 1: tcp_payload_rx, tcp_payload_bad, tcp_responses_tx
  • page 2: tcp_client_tx_ok, tcp_client_tx_fail, tcp_client_send_fail_streak_max
  • page 3: tcp_server_recreate_count, tcp_client_recreate_count, tcp_client_reconnect_calls
  • page 4: reason_last, reason_arg_last, reason_update_count
  • page 5: server_burst_requested_max, client_burst_requested_max, tcp_client_not_connected_ticks

Current reason codes:

  • 0: none/reset
  • 1: boot
  • 2: force-disconnect command
  • 3: server recreate
  • 4: client recreate by command
  • 5: client recreate watchdog
  • 6: client reconnect poll
  • 7: client send fail streak started

6. Quality Gate

Strict matrix (base + aggressive):

./tools/example_tcpip_quality_gate.sh \
  --board-ip 192.168.1.7 \
  --host-bind 192.168.1.9 \
  --base-runs 20 \
  --aggr-runs 5 \
  --health-at-end

This script stores all logs in out/quality-gate/<timestamp>/. Useful flags:

  • --health-pages 6
  • --health-at-end
  • --stop-on-first-fail

7. Soak Test

Long-running soak with automatic pass/fail summary:

./tools/example_tcpip_soak.sh \
  --board-ip 192.168.1.7 \
  --host-bind 192.168.1.9 \
  --duration-min 480 \
  --strict-client-stream \
  --health-pages 6 \
  --max-failures 1

This script stores all logs in out/soak/<timestamp>/.

8. Multi-Hour / Overnight Soak

Use the long-run wrapper to execute for hours and get final pass ratio + fail breakdown:

./tools/example_tcpip_soak_hours.sh \
  --board-ip 192.168.1.7 \
  --host-bind 192.168.1.9 \
  --hours 8 \
  --min-pass-ratio 0.90 \
  --baseline-pass-ratio 0.8475

Outputs:

  • Run console/session log: out/soak-hours/<timestamp>.log
  • Per-run logs from soak engine: out/soak/<timestamp>/

To leave it running in background:

nohup ./tools/example_tcpip_soak_hours.sh --board-ip 192.168.1.7 --host-bind 192.168.1.9 --hours 8 > out/soak-hours/latest.nohup.log 2>&1 &