Initial commit of DPU Service generation code by aadvani-nvidia · Pull Request #814 · NVIDIA/ncx-infra-controller-core

aadvani-nvidia · 2026-04-04T02:02:40Z

Description

Adds code for creating DPF configuration files for dhcp, doca-hbn and dpu-agent

Type of Change

[ x] Add - New feature or capability
Change - Changes in existing functionality
Fix - Bug fixes
Remove - Removed features or deprecated functionality
Internal - Internal changes (refactoring, tests, docs, etc.)

Related Issues (Optional)

Breaking Changes

This PR contains breaking changes

Testing

Unit tests added/updated
Integration tests added/updated
[ x] Manual testing performed
No testing required (docs, internal refactor, etc.)

Additional Notes

Tested by generating a yaml output and comparing with existing DPF configuration files
There is still pending work to setup the chains for the DpuDeployment

github-actions · 2026-04-04T02:05:21Z

🔐 TruffleHog Secret Scan

✅ No secrets or credentials found!

Your code has been scanned for 700+ types of secrets and credentials. All clear! 🎉

🔗 View scan details

_{🕐 Last updated: 2026-04-04 02:05:21 UTC | Commit: ad4aa0e}

FrankSpitulski · 2026-04-04T02:14:29Z

crates/dpf/src/services.rs

+pub const DOCA_HBN_SERVICE_NETWORK: &str = "mybrhbn";
+
+/// DHCP Service Definitions
+pub const DHCP_SERVER_SERVICE_NAME: &str = "carbide-dhcp-server";


these don't go in the DPF SDK crate. pull the carbide configs out into the carbide DPF module with the default v2 services

Signed-off-by: Ashok Advani <aadvani@nvidia.com>

- Added DpuServiceInterface and ServiceChains Signed-off-by: Ashok Advani <aadvani@nvidia.com>

Copilot

Pull request overview

Adds initial “DPU Service generation” support to the DPF SDK/API, extending initialization to generate additional CRDs (Service NADs and Service Interfaces) and introducing Carbide-specific v2 service definitions.

Changes:

Extend SDK service types to include Service NAD specification and interface template types.
Add repositories + SDK initialization logic to create/apply DPUServiceNAD and DPUServiceInterface resources.
Update API setup to supply an explicit v2 service set; add YAML-generation helper tests.

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 9 comments.

Show a summary per file

File	Description
crates/dpf/src/types.rs	Adds Service NAD and service-interface-template SDK types.
crates/dpf/src/sdk.rs	Builds/applies NAD + service interface templates during initialization; changes deployment service-chain construction.
crates/dpf/src/repository/traits.rs	Introduces `DpuServiceNADRepository` and adds `apply()` to `DpuServiceInterfaceRepository`.
crates/dpf/src/repository/kube.rs	Implements Kube-backed CRUD/apply for `DPUServiceNAD` and `DPUServiceInterface`.
crates/dpf/src/test/sdk_initialization.rs	Extends mock repos and adds YAML-generation tests for initialized resources/v2 services.
crates/dpf/src/lib.rs	Re-exports new Service NAD SDK types.
crates/dpf/Cargo.toml	Adds `serde_yaml` dependency (currently used by tests).
crates/api/src/setup.rs	Populates v2 services list instead of an empty list.
crates/api/src/dpf_services.rs	Adds Carbide v2 service definitions and default registry/image constants.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-04-07T22:40:43Z

crates/dpf/src/types.rs

+/// Service Network Attachment Definition (NAD)
+#[derive(Debug, Clone)]
+pub enum ServiceNADResourceType {
+    Vf,
+    Sf,
+    Veth,
+}


ServiceNADResourceType is used by reference in sdk::build_service_nad, but the enum is not Copy. Matching on service_nad.resource_type will try to move the value out of &ServiceNAD and won’t compile. Fix by deriving Copy for this enum (and/or matching on &service_nad.resource_type / cloning).

Copilot · 2026-04-07T22:40:43Z

crates/dpf/src/types.rs

+/// Service Network Attachment Definition (NAD)
+#[derive(Debug, Clone)]
+pub enum DpuServiceInterfaceTemplateType {
+    Vlan,
+    Physical,
+    Pf,
+    Vf,
+    Ovn,
+    Service,
+}


DpuServiceInterfaceTemplateType is matched on via iface.iface_type in sdk::apply_service_interface_templates, but the enum is not Copy. This will attempt to move the value out of a borrowed &DpuServiceInterfaceTemplateDefinition and won’t compile. Derive Copy for this enum (and/or match on a reference / clone before matching).

Copilot · 2026-04-07T22:40:43Z

crates/dpf/src/types.rs

+}
+
+
+/// Service Network Attachment Definition (NAD)


The comment above DpuServiceInterfaceTemplateType says “Service Network Attachment Definition (NAD)”, but the enum represents service interface template types. This is misleading documentation for a public SDK type; please update the doc comment to describe interface templates instead of NADs.

Suggested change

/// Service Network Attachment Definition (NAD)

/// Interface template type for a DPU service.

Copilot · 2026-04-07T22:40:44Z

crates/dpf/src/sdk.rs

+    let all_switches: Vec<DpuDeploymentServiceChainsSwitches> = vec![
+        DpuDeploymentServiceChainsSwitches {
+            ports: vec![
+                DpuDeploymentServiceChainsSwitchesPorts {
+                    service_interface: Some(
+                        DpuDeploymentServiceChainsSwitchesPortsServiceInterface {
+                            match_labels: BTreeMap::from([(
+                                "interface".to_string(),
+                                "p0".to_string(),
+                            )]),
+                            ipam: None,
                        },
-                        DpuDeploymentServiceChainsSwitchesPorts {
-                            service: Some(DpuDeploymentServiceChainsSwitchesPortsService {
-                                name: chain.service_name.clone(),
-                                interface: chain.service_interface.clone(),
-                                ipam: None,
-                            }),
-                            service_interface: None,
+                    ),
+                    service: Some(
+                        DpuDeploymentServiceChainsSwitchesPortsService {
+                            name: "doca-hbn".to_string(),
+                            interface: "p0_if".to_string(),
+                            ipam: None,
+                        }
+                    )
+                }
+            ],
+            service_mtu: None,
+        },
+        DpuDeploymentServiceChainsSwitches {
+            ports: vec![
+                DpuDeploymentServiceChainsSwitchesPorts {
+                    service_interface: Some(
+                        DpuDeploymentServiceChainsSwitchesPortsServiceInterface {
+                            match_labels: BTreeMap::from([(
+                                "interface".to_string(),
+                                "pf0hpf".to_string(),
+                            )]),
+                            ipam: None,
                        },
-                    ],
-                    service_mtu: None,
-                })
-        })
-        .collect();
+                    ),
+                    service: Some(
+                        DpuDeploymentServiceChainsSwitchesPortsService {
+                            name: "doca-hbn".to_string(),
+                            interface: "pf0hpf_if".to_string(),
+                            ipam: None,
+                        }
+                    )
+                },
+                DpuDeploymentServiceChainsSwitchesPorts {
+                    service_interface: None,
+                    service: Some(
+                        DpuDeploymentServiceChainsSwitchesPortsService {
+                            name: "carbide-dhcp-server".to_string(),
+                            interface: "d_pf0hpf_if".to_string(),
+                            ipam: None,
+                        }
+                    )
+                }
+            ],
+            service_mtu: None,
+        }
+    ];


build_deployment() no longer derives service_chains from svc.service_chain_switches and instead hard-codes switches for specific service/interface names (e.g. doca-hbn, carbide-dhcp-server). This breaks callers that supply their own service_chain_switches and can generate incorrect deployments when services are renamed or omitted. Restore the previous behavior of building switches from services[*].service_chain_switches (or gate the hard-coded chain behind an explicit config flag).

Copilot · 2026-04-07T22:40:44Z

crates/dpf/Cargo.toml

 hyper = { features = ["client", "http1"], workspace = true }
 http = { workspace = true }
 once_cell = { workspace = true }
+serde_yaml = { workspace = true }


serde_yaml is only used in crates/dpf/src/test/sdk_initialization.rs (tests), but it is added as a normal dependency. This needlessly pulls it into non-test builds of the dpf crate. Move serde_yaml to [dev-dependencies] (or gate it behind a feature used only for these tests).

Copilot · 2026-04-07T22:40:44Z

crates/dpf/src/test/sdk_initialization.rs

+#[tokio::test]
+async fn test_generate_yaml_for_initialized_resources() {
+    let mock = InitializationMock::default();
+
+    let config = InitDpfResourcesConfig {
+        bfb_url: "http://example.com/test.bfb".to_string(),
+        deployment_name: "carbide-deployment".to_string(),
+        ..Default::default()
+    };
+
+    let sdk = crate::sdk::DpfSdkBuilder::new(mock.clone(), TEST_NS, "test-password".to_string())
+        .initialize(&config)
+        .await
+        .unwrap();
+    drop(sdk);
+
+    for entry in mock.bfbs.iter() {
+        println!("---\n{}", serde_yaml::to_string(entry.value()).unwrap());
+    }
+    for entry in mock.flavors.iter() {
+        println!("---\n{}", serde_yaml::to_string(entry.value()).unwrap());
+    }
+    for entry in mock.service_templates.iter() {
+        println!("---\n{}", serde_yaml::to_string(entry.value()).unwrap());
+    }
+    for entry in mock.service_configs.iter() {
+        println!("---\n{}", serde_yaml::to_string(entry.value()).unwrap());
+    }
+    for entry in mock.nads.iter() {
+        println!("---\n{}", serde_yaml::to_string(entry.value()).unwrap());
+    }
+    for entry in mock.deployments.iter() {
+        println!("---\n{}", serde_yaml::to_string(entry.value()).unwrap());
+    }
+}


These tests only print YAML to stdout and contain no assertions, so they won’t catch regressions and may add noisy output to CI. Consider converting them into snapshot/assertion-based tests (e.g. compare against expected YAML/JSON) or marking them #[ignore] if they’re intended as local generation helpers.

Copilot · 2026-04-07T22:40:44Z

crates/dpf/src/test/sdk_initialization.rs

+#[tokio::test]
+async fn test_generate_yaml_for_v2_services() {
+    use crate::types::{ServiceInterface, ServiceNAD, ServiceNADResourceType};
+
+    let mock = InitializationMock::default();
+
+    let doca_helm = "https://helm.ngc.nvidia.com/nvidia/doca";
+    let carbide_helm = "https://gitlab-master.nvidia.com/aadvani/my-helm-project/-/raw/main/charts-repo";
+    let doca_image = "nvcr.io/nvidia/doca";
+    let carbide_image = "https://gitlab-master.nvidia.com/aadvani/my-helm-project";
+
+    let services = vec![
+        crate::services::dts_service(&crate::services::ServiceRegistryConfig::default()),
+        ServiceDefinition {
+            helm_values: Some(serde_json::json!({
+                "image": {
+                    "repository": format!("{}/forge-dhcp-server", carbide_image),
+                    "tag": "v1.9.5-arm64-distroless",
+                }
+            })),
+            interfaces: vec![ServiceInterface {
+                name: "d_pf0hpf_if".to_string(),
+                network: "mybrsfc-dhcp".to_string(),
+            }],
+            service_nad: Some(ServiceNAD {
+                name: "mybrsfc-dhcp".to_string(),
+                bridge: Some("br-sfc".to_string()),
+                resource_type: ServiceNADResourceType::Sf,
+                ipam: Some(false),
+                mtu: Some(1500),
+            }),
+            ..ServiceDefinition::new("carbide-dhcp-server", carbide_helm, "carbide-dhcp-server", "2.0.9")
+        },
+        ServiceDefinition {
+            helm_values: Some(serde_json::json!({
+                "image": {
+                    "repository": format!("{}/doca-hbn", doca_image),
+                    "tag": "3.2.1-doca3.2.1",
+                },
+                "resources": { "memory": "6Gi", "nvidia.com/bf_sf": 2 },
+            })),
+            config_values: Some(serde_json::json!({
+                "configuration": {
+                    "startupYAMLJ2": concat!(
+                        "- header:\n",
+                        "    model: BLUEFIELD\n",
+                        "    nvue-api-version: nvue_v1\n",
+                        "    rev-id: 1.0\n",
+                        "    version: HBN 2.4.0\n",
+                        "- set:\n",
+                        "    interface:\n",
+                        "      p0_if:\n",
+                        "        type: swp\n",
+                        "      pf0hpf_if:\n",
+                        "        type: swp\n",
+                    )
+                }
+            })),
+            interfaces: vec![
+                ServiceInterface { name: "p0_if".to_string(), network: "mybrhbn".to_string() },
+                ServiceInterface { name: "pf0hpf_if".to_string(), network: "mybrhbn".to_string() },
+            ],
+            ..ServiceDefinition::new("doca-hbn", doca_helm, "doca-hbn", "1.0.5")
+        },
+        ServiceDefinition {
+            helm_values: Some(serde_json::json!({
+                "image": {
+                    "repository": format!("{}/forge-dpu-agent", carbide_image),
+                    "tag": "v0.3-arm64-multistage",
+                }
+            })),
+            ..ServiceDefinition::new("carbide-dpu-agent", carbide_helm, "carbide-dpu-agent", "0.4.0")
+        },
+    ];
+
+    let config = InitDpfResourcesConfig {
+        bfb_url: "http://example.com/test.bfb".to_string(),
+        deployment_name: "carbide-deployment".to_string(),
+        services,
+        ..Default::default()
+    };
+
+    let sdk = crate::sdk::DpfSdkBuilder::new(mock.clone(), TEST_NS, "test-password".to_string())
+        .initialize(&config)
+        .await
+        .unwrap();
+    drop(sdk);
+
+    for entry in mock.bfbs.iter() {
+        println!("---\n{}", serde_yaml::to_string(entry.value()).unwrap());
+    }
+    for entry in mock.flavors.iter() {
+        println!("---\n{}", serde_yaml::to_string(entry.value()).unwrap());
+    }
+    for entry in mock.service_templates.iter() {
+        println!("---\n{}", serde_yaml::to_string(entry.value()).unwrap());
+    }
+    for entry in mock.service_configs.iter() {
+        println!("---\n{}", serde_yaml::to_string(entry.value()).unwrap());
+    }
+    for entry in mock.nads.iter() {
+        println!("---\n{}", serde_yaml::to_string(entry.value()).unwrap());
+    }
+    for entry in mock.service_interfaces.iter() {
+        println!("---\n{}", serde_yaml::to_string(entry.value()).unwrap());
+    }
+    for entry in mock.deployments.iter() {
+        println!("---\n{}", serde_yaml::to_string(entry.value()).unwrap());
+    }
+}


This test builds a large “v2 services” config and then only prints generated YAML without assertions. If this is meant to validate generated CRDs, add concrete checks (counts, key fields, or snapshots). If it’s only a manual YAML generator, mark it #[ignore] to avoid running in normal test suites.

Copilot · 2026-04-07T22:40:45Z

crates/api/src/dpf_services.rs

+pub const DEFAULT_CARBIDE_HELM_REGISTRY: &str = "https://gitlab-master.nvidia.com/aadvani/my-helm-project/-/raw/main/charts-repo";
+
+/// Default DOCA container image registry prefix.
+pub const DEFAULT_DOCA_IMAGE_REGISTRY: &str = "nvcr.io/nvidia/doca";
+
+/// Default Carbide container image registry prefix.
+pub const DEFAULT_CARBIDE_IMAGE_REGISTRY: &str = "https://gitlab-master.nvidia.com/aadvani/my-helm-project";


The default Carbide helm/image registries are set to a specific internal GitLab project URL. As defaults, this can break deployments outside that environment and makes the API behavior environment-specific. Consider defaulting to the public/official registries (or leaving defaults empty) and relying on configuration to override for internal development.

Suggested change

pub const DEFAULT_CARBIDE_HELM_REGISTRY: &str = "https://gitlab-master.nvidia.com/aadvani/my-helm-project/-/raw/main/charts-repo";

/// Default DOCA container image registry prefix.

pub const DEFAULT_DOCA_IMAGE_REGISTRY: &str = "nvcr.io/nvidia/doca";

/// Default Carbide container image registry prefix.

pub const DEFAULT_CARBIDE_IMAGE_REGISTRY: &str = "https://gitlab-master.nvidia.com/aadvani/my-helm-project";

/// Default Carbide helm registry is intentionally unset and should be provided via configuration.

pub const DEFAULT_CARBIDE_HELM_REGISTRY: &str = "";

/// Default DOCA container image registry prefix.

pub const DEFAULT_DOCA_IMAGE_REGISTRY: &str = "nvcr.io/nvidia/doca";

/// Default Carbide container image registry prefix is intentionally unset and should be provided via configuration.

pub const DEFAULT_CARBIDE_IMAGE_REGISTRY: &str = "";

Copilot · 2026-04-07T22:40:45Z

crates/api/src/dpf_services.rs

+pub const DEFAULT_DOCA_IMAGE_REGISTRY: &str = "nvcr.io/nvidia/doca";
+
+/// Default Carbide container image registry prefix.
+pub const DEFAULT_CARBIDE_IMAGE_REGISTRY: &str = "https://gitlab-master.nvidia.com/aadvani/my-helm-project";


DEFAULT_CARBIDE_IMAGE_REGISTRY includes an https:// scheme, but container image references typically must be in the form registry/namespace without a URL scheme. If this value is used to build the Helm image.repository, it will likely produce invalid image references (e.g. https://.../forge-dpu-agent). Consider storing the registry as gitlab-master.nvidia.com/... (no scheme) and keeping transport/protocol out of the image name.

Suggested change

pub const DEFAULT_CARBIDE_IMAGE_REGISTRY: &str = "https://gitlab-master.nvidia.com/aadvani/my-helm-project";

pub const DEFAULT_CARBIDE_IMAGE_REGISTRY: &str = "gitlab-master.nvidia.com/aadvani/my-helm-project";

Signed-off-by: Ashok Advani <aadvani@nvidia.com>

aadvani-nvidia requested a review from a team as a code owner April 4, 2026 02:02

aadvani-nvidia requested review from FrankSpitulski, abvarshney-nv and Copilot April 4, 2026 02:02

Copilot started reviewing on behalf of aadvani-nvidia April 4, 2026 02:03 View session

aadvani-nvidia review requested due to automatic review settings April 4, 2026 02:04

FrankSpitulski requested changes Apr 4, 2026

View reviewed changes

aadvani-nvidia added 2 commits April 7, 2026 15:35

Initial commit of DPU Service generation code

b048471

Signed-off-by: Ashok Advani <aadvani@nvidia.com>

- Moved dpf service template generation to dpf_services.c

5f8e370

- Added DpuServiceInterface and ServiceChains Signed-off-by: Ashok Advani <aadvani@nvidia.com>

aadvani-nvidia force-pushed the dpf-services branch from ad4aa0e to 5f8e370 Compare April 7, 2026 22:36

Copilot AI review requested due to automatic review settings April 7, 2026 22:36

Copilot started reviewing on behalf of aadvani-nvidia April 7, 2026 22:36 View session

Copilot AI reviewed Apr 7, 2026

View reviewed changes

aadvani-nvidia added 2 commits April 7, 2026 17:00

- Lint cleanup

f8daa02

Signed-off-by: Ashok Advani <aadvani@nvidia.com>

Fix DPUDeployment Service Chain

1cff033

Signed-off-by: Ashok Advani <aadvani@nvidia.com>

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Initial commit of DPU Service generation code#814

Initial commit of DPU Service generation code#814
aadvani-nvidia wants to merge 4 commits intoNVIDIA:mainfrom
aadvani-nvidia:dpf-services

aadvani-nvidia commented Apr 4, 2026

Uh oh!

github-actions bot commented Apr 4, 2026

Uh oh!

FrankSpitulski Apr 4, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 7, 2026

Uh oh!

Copilot AI Apr 7, 2026

Uh oh!

Copilot AI Apr 7, 2026

Uh oh!

Copilot AI Apr 7, 2026

Uh oh!

Copilot AI Apr 7, 2026

Uh oh!

Copilot AI Apr 7, 2026

Uh oh!

Copilot AI Apr 7, 2026

Uh oh!

Copilot AI Apr 7, 2026

Uh oh!

Copilot AI Apr 7, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

	/// Service Network Attachment Definition (NAD)
	/// Interface template type for a DPU service.

	pub const DEFAULT_CARBIDE_IMAGE_REGISTRY: &str = "https://gitlab-master.nvidia.com/aadvani/my-helm-project";
	pub const DEFAULT_CARBIDE_IMAGE_REGISTRY: &str = "gitlab-master.nvidia.com/aadvani/my-helm-project";

Conversation

aadvani-nvidia commented Apr 4, 2026

Description

Type of Change

Related Issues (Optional)

Breaking Changes

Testing

Additional Notes

Uh oh!

github-actions bot commented Apr 4, 2026

🔐 TruffleHog Secret Scan

Uh oh!

FrankSpitulski Apr 4, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants