Merge pull request #20 from tenstorrent/sbansal/check_eth_fw_ver

Added fix to make local and remote info agnostic to eth fw update changes

CHANGELOG.md

@@ -5,6 +5,14 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## 1.2.0 - 06/03/2025
+
+### Fixed
+
+- Updated how local eth board info is calculated to make it agnostic to eth fw version
+- bumped tt-tools-common version
+- Added traceback printing when catching exceptions in main.
+
 ## 1.1.5 - 14/05/2024
 
 ### Updated

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "tt-topology"
-version = "1.1.5"
+version = "1.2.0"
 description = "ethernet topology configuration tool for Tenstorrent silicon"
 readme = "README.md"
 requires-python = ">=3.7"
@@ -28,7 +28,7 @@ dependencies = [
   'elasticsearch==8.11.0',
   'pydantic>=1.2',
   'pyluwen @ git+https://github.com/tenstorrent/luwen.git@v0.4.9#subdirectory=crates/pyluwen',
-  'tt_tools_common @ git+https://github.com/tenstorrent/tt-tools-common.git@v1.4.11',
+  'tt_tools_common @ git+https://github.com/tenstorrent/tt-tools-common.git@v1.4.14',
   'pre-commit==3.5.0',
   'networkx==3.1',
   'matplotlib==3.7.4'

tt_topology/backend.py

@@ -206,7 +206,7 @@ def get_eth_config_state(self):
         for data in config_state:
             assert (
                 data["fw_version"] == config_state[0]["fw_version"]
-            ), "Firmware versions do not match"
+            ), f"Firmware versions do not match: {data['fw_version']} != {config_state[0]['fw_version']}"
         if not self.log.starting_configs:
             self.log.starting_configs = config_state_log
         elif not self.log.post_default_flashing_configs:
@@ -296,6 +296,24 @@ def flash_to_default_state(self):
                 CMD_LINE_COLOR.ENDC,
             )
 
+    def get_local_eth_board_info(self, chip):
+        """
+        Get the local board info from noc, making it eth fw version agnostic
+        """
+        local_board_id = bytearray(4)
+        local_board_type = bytearray(4)
+
+        for port in range(16):
+            eth_x, eth_y = self.eth_xy_decode(port)
+            chip.noc_read(0, eth_x, eth_y, constants.ETH_TEST_RESULT_LOCAL_TYPE, local_board_type)
+
+            if int.from_bytes(local_board_type, "little") != 0:
+                chip.noc_read(0, eth_x, eth_y, constants.ETH_TEST_RESULT_LOCAL_ID, local_board_id)
+                local_board_type = int.from_bytes(local_board_type, "little")
+                local_board_id = int.from_bytes(local_board_id, "little")
+                local_board_info = f"{(local_board_type << 32) | local_board_id:016x}"
+                return local_board_info
+
     def generate_connection_map(self):
         """
         Generate an map with chip data and a list of connections
@@ -314,15 +332,7 @@ def generate_connection_map(self):
             chip = device.as_wh()
             board_id = str(hex(device.board_id())).replace("0x", "")
             board_type = get_board_type(board_id)
-
-            # print(board_id, board_type)
-            eth_board_type = bytearray(4)
-            eth_board_id = bytearray(4)
-            chip.noc_read(0, 9, 0, constants.ETH_L1_PARAM_BOARD_TYPE, eth_board_type)
-            chip.noc_read(0, 9, 0, constants.ETH_L1_PARAM_BOARD_ID, eth_board_id)
-            eth_board_type = int.from_bytes(eth_board_type, "little")
-            eth_board_id = int.from_bytes(eth_board_id, "little")
-            eth_board_info = f"{(eth_board_type << 32) | eth_board_id:016x}"
+            eth_board_info = self.get_local_eth_board_info(chip)
 
             chip_data[eth_board_info] = {
                 "id": idx,
@@ -357,6 +367,10 @@ def generate_connection_map(self):
                     connection_map_log_obj = log_obj
                     break
 
+            # get fw version and collect remote_info accordingly
+            chip_eth_fw_ver = bytearray(4)
+            chip.spi_read(int(constants.ETH_FW_VERSION_ADDR), chip_eth_fw_ver)
+            chip_eth_fw_ver = int.from_bytes(chip_eth_fw_ver, "little")
             # Go through all 16 ETH ports and read their remote chip ids (if applicable)
             # Use those IDs to construct the vectorized representation
             for port in range(16):

tt_topology/constants.py

@@ -25,3 +25,8 @@
 ETH_TEST_RESULT_REMOTE_ID           = ETH_TEST_RESULT_BASE_ADDR + (0x4) * 73
 ETH_TEST_RESULT_REMOTE_COORD        = ETH_TEST_RESULT_BASE_ADDR + (0x4) * 74  # 0x0000YYXX
 ETH_TEST_RESULT_REMOTE_SHELF_RACK   = ETH_TEST_RESULT_BASE_ADDR + (0x4) * 75  # 0x0000SSRR
+
+ETH_TEST_RESULT_LOCAL_TYPE         = ETH_TEST_RESULT_BASE_ADDR + (0x4) * 64
+ETH_TEST_RESULT_LOCAL_ID           = ETH_TEST_RESULT_BASE_ADDR + (0x4) * 65
+ETH_TEST_RESULT_LOCAL_COORD        = ETH_TEST_RESULT_BASE_ADDR + (0x4) * 66  # 0x0000YYXX
+ETH_TEST_RESULT_LOCAL_SHELF_RACK   = ETH_TEST_RESULT_BASE_ADDR + (0x4) * 67  # 0x0000SSRR

tt_topology/tt_topology.py

@@ -9,6 +9,7 @@
 import sys
 import time
 import argparse
+import traceback
 import pkg_resources
 from tt_tools_common.reset_common.wh_reset import WHChipReset
 from tt_tools_common.ui_common.themes import CMD_LINE_COLOR
@@ -467,10 +468,10 @@ def main():
     except Exception as e:
         print(
             CMD_LINE_COLOR.RED,
-            e,
+            traceback.format_exc(),
             CMD_LINE_COLOR.ENDC,
         )
-        topo_backend.log.errors = str(e)
+        topo_backend.log.errors = str(traceback.format_exc())
         errors = True
     finally:
         # Still collect the log if something went wrong