[Nvidia-bluefield] Support for Standalone DPU in nvidia bluefield platform by gpunathilell · Pull Request #56 · gpunathilell/sonic-buildimage

gpunathilell · 2025-12-01T17:25:53Z

Why I did it

Support for new DPU platform and for the Standalone DPU "Appliance" Topology
Changes introduced:
Platform independent changes:

dockers/docker-database/docker-database-init.sh Support for Appliance DPU, do not export REMOTE_DB_PORT and REMOTE_DB_IP in case of appliance DPU, so that database is initialized with local DPU_APPL_DB and DPU_STATE_DB in case of appliance topology
dockers/docker-orchagent/orchagent.sh - Allow zmq initialization in orchagent for Appliance subtype
dockers/docker-sonic-gnmi/gnmi-native.sh - Allow zmq port initialization in the gnmi server for Appliance subtype
src/sonic-config-engine/config_samples.py - Add support for appliance topology and generate basic default config during first boot
src/sonic-yang-models/yang-models/sonic-device_metadata.yang - Yang validation update for appliance topology
src/systemd-sonic-generator/systemd-sonic-generator.cpp - Differentiate between APPLIANCE_DPU and SMARTSWITCH_DPU during systemd service generation, this is done by checking for APPLIANCE_DPU key in platform.json
---------Platform Specific changes-------------------
platform/nvidia-bluefield/installer/install.sh.j2 and platform/nvidia-bluefield/installer/create_sonic_image Support in bfb image generation for Appliance DPU (Generate the right grub configuration for CPU usage for NASA in case of 8 core vs 16 core device and execute reboot from the DPU in case of standalone DPU instead of waiting for host reboot)
platform/nvidia-bluefield/bluefield-platform-modules/udev/82-net-setup-link.rules Restore creation of 2 dataplane ports (Although at this point only one port is supported in NASA)
platform/nvidia-bluefield/platform-api/sonic_platform/chassis.py and platform/nvidia-bluefield/platform-api/sonic_platform/device_data.py - Return False for is_smartswitch check, used in thermalctld to push data to smartswitch (Not needed for appliance)
device/nvidia-bluefield/arm64-nvda_bf-9009d3b400ccea/Nvidia-9009d3b400CCEA - New DPU platform for 8-core Bluefield 3 Device - Nvidia-9009d3b400CCEA Including all Platform specific files and port configuration
device/nvidia-bluefield/arm64-nvda_bf-9009d3b600cvaa/platform.json - Add APPLIANCE_DPU key to detect whether platform is an appliance DPU or Smartswitch DPU

Work item tracking

Microsoft ADO (number only):

How I did it

Updated changes as described above

How to verify it

Test with manual deployment of image generated after build on standalone DPU device to test the SONiC image

Which release branch to backport (provide reason below if selected)

Tested branch (Please provide the tested image version)

Description for the changelog

Link to config_db schema for YANG module changes

A picture of a cute animal (not mandatory but encouraged)

oleksandrivantsiv · 2025-12-01T17:28:35Z

+{
+    "interfaces": {
+        "Ethernet0": {
+            "default_brkout_mode": "1x100G"


Shouldn't it be 1x200G?

oleksandrivantsiv · 2025-12-01T17:29:00Z

+SAI_DUMP_STORE_PATH=/var/log/bluefield/sdk-dumps
+SAI_DUMP_STORE_AMOUNT=10
+DASH_ACL_DEFAULT_RULE_ACTION=permit
+PORT_LAYOUT=2x100G


oleksandrivantsiv · 2025-12-01T17:29:24Z

+            "index": "1,1,1,1",
+            "lanes": "0,1,2,3",
+            "breakout_modes": {
+                "1x100G": ["etp1"]


oleksandrivantsiv · 2025-12-01T17:30:34Z

-    midplane_ip=$( ip -4 -o addr show eth0-midplane | awk '{print $4}' | cut -d'/' -f1  )
-    if [[ $midplane_ip != "" ]]
-    then
+    if [[ $IS_APPLIANCE_DPU == "false" ]]; then


This should be in separate PR

Split into separate PR

oleksandrivantsiv · 2025-12-01T17:31:22Z

 # Enable ZMQ
 LOCALHOST_SUBTYPE=`sonic-db-cli CONFIG_DB hget "DEVICE_METADATA|localhost" "subtype"`
-if [[ x"${LOCALHOST_SUBTYPE}" == x"SmartSwitch" ]]; then
+if [[ x"${LOCALHOST_SUBTYPE}" == x"SmartSwitch" || x"${LOCALHOST_SUBTYPE}" == x"Appliance" ]]; then


Please put in separate PR all related to the vendor independent implementation

oleksandrivantsiv · 2025-12-08T20:50:11Z

@@ -0,0 +1,74 @@
+#!/bin/bash
+#
+# Copyright (c) 2024 NVIDIA CORPORATION & AFFILIATES.


oleksandrivantsiv · 2025-12-08T20:50:50Z

+    platform=$(sonic-cfggen -d -v 'DEVICE_METADATA["localhost"]["platform"]')
+    single_port=$([[ $hwsku == *"-com-dpu" ]] && echo true || echo false)
+    # Read MAC addresses
+    base_mac="$(echo $SYNCD_VARS | jq -r '.mac')"


This is duplication. Please remove

oleksandrivantsiv · 2025-12-08T20:53:14Z

+
+    if [[ $platform == "arm64-nvda_bf-9009d3b400ccea" ]]; then
+        # 8 cores DPU, reduce hugepages to 4096
+        echo 4096 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages


let's have humber of huge pages defined in the device data directory.

please add new key/value pair to platform.json

oleksandrivantsiv · 2025-12-08T20:54:46Z

 declare -r FORCE_FW_CONFIG_RESET="${FORCE_FW_CONFIG_RESET:-false}"

+# List of platforms that should skip firmware upgrade
+PLATFORMS_SKIP_FW_UPGRADE=(
+	"9009d3b600"


let's define new flag in platform.json

oleksandrivantsiv · 2025-12-11T00:01:38Z

@@ -32,5 +32,6 @@
                "1x200G": ["etp2"]
            }
        }
-    }
+    },
+    "SKIP_FW_UPDATE": ""


this should be the same as arm64-nvda_bf-9009d3b600cvaa

oleksandrivantsiv · 2025-12-11T00:03:28Z

@@ -46,6 +46,7 @@ if [[ -n "$lspci_output" ]]; then
    for key in "${DEVICE_ORDER[@]}"; do
        if echo "$lspci_output" | grep "$VENDOR_ID:$key" &>/dev/null; then
            DEVICE_TYPE="${DEVICE_DICT[$key]}"
+            DEVICE_PCI_ID=$(echo "$lspci_output" | grep "$VENDOR_ID:$key" | awk '{print $1}' | head -n 1)


Where is this variable used?

It wasnt clear because master wasnt up to date to RC, please check now, variable is used

oleksandrivantsiv · 2025-12-11T00:03:57Z

 declare -r FORCE_FW_CONFIG_RESET="${FORCE_FW_CONFIG_RESET:-false}"

+


Please remove redundant lines

oleksandrivantsiv · 2026-01-12T21:30:26Z

Please update the header date in all modified files

oleksandrivantsiv · 2026-01-12T21:41:38Z

-	sleep 1
-done
+
+if [[ $SELF_REBOOT == "false" ]]; then


SELF_REBOOT != "true", this will be more "bash" like way to handle

Signed-off-by: gpunathilell <gpunathilell@nvidia.com>

…tically (sonic-net#25545) #### Why I did it src/sonic-dash-api ``` * 459cad6 - (HEAD -> master, origin/master, origin/HEAD) Add two extra HA states (#59) (3 hours ago) [Changrong Wu] * b3dfda7 - Add list of trusted VNIs (#56) (29 hours ago) [Lawrence Lee] ``` #### How I did it #### How to verify it #### Description for the changelog

gpunathilell requested a review from oleksandrivantsiv December 1, 2025 17:25

gpunathilell requested a review from dgsudharsan as a code owner December 1, 2025 17:25

oleksandrivantsiv reviewed Dec 1, 2025

View reviewed changes

Comment thread device/nvidia-bluefield/arm64-nvda_bf-9009d3b400ccea/system_health_monitoring_config.json

oleksandrivantsiv reviewed Dec 1, 2025

View reviewed changes

gpunathilell force-pushed the BF3_POC branch from 2d29f6a to 5044430 Compare December 2, 2025 16:51

oleksandrivantsiv reviewed Dec 8, 2025

View reviewed changes

oleksandrivantsiv reviewed Dec 11, 2025

View reviewed changes

oleksandrivantsiv reviewed Jan 12, 2026

View reviewed changes

oleksandrivantsiv approved these changes Jan 12, 2026

View reviewed changes

Support for standalone DPU in Nvidia Bluefield platform

28b2105

Signed-off-by: gpunathilell <gpunathilell@nvidia.com>

gpunathilell force-pushed the BF3_POC branch from 87d62e6 to 28b2105 Compare January 15, 2026 01:34

dgsudharsan removed their request for review April 1, 2026 03:56

Merge branch 'master' into BF3_POC

72f93d2

		declare -r FORCE_FW_CONFIG_RESET="${FORCE_FW_CONFIG_RESET:-false}"

Conversation

gpunathilell commented Dec 1, 2025

Why I did it

Work item tracking

How I did it

How to verify it

Which release branch to backport (provide reason below if selected)

Tested branch (Please provide the tested image version)

Description for the changelog

Link to config_db schema for YANG module changes

A picture of a cute animal (not mandatory but encouraged)

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

oleksandrivantsiv commented Jan 12, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants