dataplane init system part 1 #897

daniel-noland · 2025-10-09T01:34:11Z

This is a first installment of the init package.

It will eventually be responsible for starting the dataplane, but is not currently and will not be even if this is merged.

That will require a few extra steps.

As such, this code, merged or not, will have no impact on the rest of the development for this workspace just yet. I'm only submitting this part early to try to keep the PRs as byte size and manageable chunks.

Copilot

Pull Request Overview

This pull request introduces the foundation for a dataplane initialization system with new NIC management capabilities. The changes focus on creating infrastructure for binding network cards to appropriate drivers and setting up the system initialization flow.

Key changes:

Created a new init crate with PCI device and driver management functionality
Added comprehensive documentation and configuration for the initialization system
Updated various crates to use embedded README documentation and improved error handling

Reviewed Changes

Copilot reviewed 35 out of 39 changed files in this pull request and generated 7 comments.

Show a summary per file

File	Description
`init/`	New crate for dataplane initialization with PCI device management
`the-rules.md`	Comprehensive documentation of DPDK + async threading rules
`start-up.mmd`	Mermaid diagram for proposed startup sequence
`tracectl/`	Updated to use embedded README documentation
`dpdk/`	Improved error handling and memory management
`dataplane/`	Disabled some components and added build dependencies
Various config files	Updated dependencies, build settings, and documentation

tracectl/README.md

init/src/sysfs.rs

init/src/main.rs

dpdk/src/mem.rs

dpdk/src/eal.rs

Copilot

Pull Request Overview

Copilot reviewed 35 out of 39 changed files in this pull request and generated no new comments.

The init package (more properly `dataplane-init`), is responsible for starting up the dataplane and forwarding computed or augmented command line arguments to it. This commit only includes basic documentation and mission statement. This commit series focuses on initializing the network card in a manor suitable for use by DPDK. Signed-off-by: Daniel Noland <[email protected]>

Copilot

Pull Request Overview

Copilot reviewed 8 out of 9 changed files in this pull request and generated 4 comments.

Comments suppressed due to low confidence (1)

init/src/nic.rs:1

Redundant error handling: the ? operator is unnecessary after Err(...) since Err is already being returned. Remove the ? operator.

// SPDX-License-Identifier: Apache-2.0

init/src/sysfs.rs

init/src/nic.rs

Copilot

Pull Request Overview

Copilot reviewed 8 out of 9 changed files in this pull request and generated 2 comments.

Comments suppressed due to low confidence (1)

init/src/nic.rs:1

The ? operator is unnecessary here since this is already the final expression in the match arm. The expression should be simplified to just Err(InitErr::IoError(e.into())).

// SPDX-License-Identifier: Apache-2.0

init/src/nic.rs

init/src/main.rs

Introduce a package wide application error type. This is very different from our more typical library level error types in that it is intended to be a catch-all error type for fatal errors. Almost nothing which can go wrong in this crate is expected to be recoverable without human intervention. Signed-off-by: Daniel Noland <[email protected]>

This will require extension to customize log level. Signed-off-by: Daniel Noland <[email protected]>

Copilot

Pull Request Overview

Copilot reviewed 8 out of 9 changed files in this pull request and generated 3 comments.

Copilot · 2025-10-10T01:46:04Z

init/src/nic.rs

+    fn driver(&self) -> Result<Option<PciDriver>, InitErr> {
+        let device_path = self.device_path()?;
+        info!("found device {self} under device path {:?}", device_path);
+        let driver_path = device_path.relative("driver")?;


The code attempts to get the driver path but doesn't handle the case where the device has no driver bound (symlink doesn't exist). This will cause relative() to fail with an IO error instead of returning Ok(None) as the function signature suggests.

true. Will fix in the next PR

init/src/sysfs.rs

init/src/main.rs

Basic guard rails which discourage us from making some mistakes. Features: - SysfsPath type ensuring paths are under sysfs mount - UTF-8 validation to prevent log injection attacks - Filesystem type verification using statfs - Path canonicalization to handle symlinks safely Signed-off-by: Daniel Noland <[email protected]>

This function finds our sysfs mount and panics if it is missing. It also sets up tracing. Signed-off-by: Daniel Noland <[email protected]>

Introduces PCI NIC abstraction for driver management with: - Driver detection and identification - Unbinding from kernel drivers - Binding to vfio-pci for DPDK - Support for i40e, mlx5 (partial), virtio-net, and vfio-pci drivers Implements traits for safe driver manipulation through sysfs. Signed-off-by: Daniel Noland <[email protected]>

Implements basic NIC driver binding for DPDK compatibility. Binds PCI network devices to vfio-pci driver as required by DPDK for device control. This is part 1 of the dataplane initialization system, focusing on minimal viable functionality for team testing. Future work: - Add proper argument parsing - Implement privilege dropping (especially CAP_SYS_ADMIN) - Add exec to dataplane process after success Signed-off-by: Daniel Noland <[email protected]>

qmonnet

I've got concerns about the name for the crate. init sounds like we initialise the dataplane itself, not like a preliminary step to set up the NICs etc. before launching the dataplane executable. I also fear it might lead to confusion (if not for us, for users/costumers) with the init process on some Linux distribs. How about bootstrap, maybe?

Great work otherwise!

qmonnet · 2025-10-10T09:17:48Z

init/src/main.rs

+#![doc = include_str!("../README.md")]
+#![deny(clippy::pedantic, missing_docs)]
+
+fn main() {}


Your process will be named init (unless I'm mistaken). I think it would maybe make sense to find another name instead to avoid any confusion with the init process on some Linux distributions. Even if it might be clear to us in the context, it will become a source of confusion when debugging customer issues if we mention an init process.

Perhaps hwinit is a better name?

qmonnet · 2025-10-10T09:19:43Z

init/src/main.rs

+    /// Some [`std::io::Error`] error occurred
+    #[error(transparent)]
+    IoError(#[from] std::io::Error),
+    /// Invalid UTF-8 in a path under sysfs is an absolutely wild error case I expect to


This is a rustdoc comment for API docs. Who is “I”? 🙂

Suggested change

/// Invalid UTF-8 in a path under sysfs is an absolutely wild error case I expect to

/// Invalid UTF-8 in a path under sysfs is an absolutely wild error case we expect to

qmonnet · 2025-10-10T09:23:22Z

init/src/nic.rs

+///
+/// Note that the NIC may or may not be visible to the OS, depending on the state of
+/// the system.
+pub struct PciNic {


I can't help but read this “PicNic” every time I see it 😎

daniel-noland force-pushed the pr/daniel-noland/dpdk-back branch from 7658bd3 to 4b3877d Compare October 9, 2025 01:37

daniel-noland requested a review from Copilot October 9, 2025 01:38

daniel-noland self-assigned this Oct 9, 2025

daniel-noland added the enhancement New feature or request label Oct 9, 2025

Copilot AI reviewed Oct 9, 2025

View reviewed changes

daniel-noland force-pushed the pr/daniel-noland/dpdk-back branch from 4b3877d to 43ffbfe Compare October 9, 2025 01:39

daniel-noland requested a review from Copilot October 9, 2025 01:41

Copilot AI reviewed Oct 9, 2025

View reviewed changes

daniel-noland force-pushed the pr/daniel-noland/dpdk-back branch 2 times, most recently from a2b236f to 77bdba9 Compare October 9, 2025 02:05

daniel-noland changed the title ~~dataplane init system~~ dataplane init system part 1 Oct 9, 2025

daniel-noland force-pushed the pr/daniel-noland/dpdk-back branch 2 times, most recently from dba8f78 to 11c261e Compare October 9, 2025 05:00

daniel-noland marked this pull request as ready for review October 9, 2025 05:07

daniel-noland requested a review from a team as a code owner October 9, 2025 05:07

daniel-noland requested review from Copilot and sergeymatov and removed request for a team October 9, 2025 05:07

Copilot AI reviewed Oct 9, 2025

View reviewed changes

init/src/sysfs.rs Outdated Show resolved Hide resolved

init/src/sysfs.rs Outdated Show resolved Hide resolved

init/src/nic.rs Outdated Show resolved Hide resolved

init/src/nic.rs Outdated Show resolved Hide resolved

daniel-noland force-pushed the pr/daniel-noland/dpdk-back branch from 11c261e to 96b6a25 Compare October 9, 2025 05:18

daniel-noland requested a review from Copilot October 9, 2025 05:21

Copilot AI reviewed Oct 9, 2025

View reviewed changes

init/src/nic.rs Outdated Show resolved Hide resolved

init/src/main.rs Outdated Show resolved Hide resolved

daniel-noland force-pushed the pr/daniel-noland/dpdk-back branch 2 times, most recently from 45f5d60 to 51bc4f9 Compare October 9, 2025 05:25

daniel-noland added 2 commits October 9, 2025 05:37

feat(init): add basic tracing

2456016

This will require extension to customize log level. Signed-off-by: Daniel Noland <[email protected]>

daniel-noland force-pushed the pr/daniel-noland/dpdk-back branch 3 times, most recently from d9b6290 to 5803d40 Compare October 9, 2025 05:54

daniel-noland requested a review from Copilot October 10, 2025 01:45

Copilot AI reviewed Oct 10, 2025

View reviewed changes

daniel-noland added 4 commits October 10, 2025 01:51

feat(init): introduce setup function

0b47045

This function finds our sysfs mount and panics if it is missing. It also sets up tracing. Signed-off-by: Daniel Noland <[email protected]>

daniel-noland force-pushed the pr/daniel-noland/dpdk-back branch from 5803d40 to 0b5aad6 Compare October 10, 2025 01:52

qmonnet requested changes Oct 10, 2025

View reviewed changes

	/// Invalid UTF-8 in a path under sysfs is an absolutely wild error case I expect to
	/// Invalid UTF-8 in a path under sysfs is an absolutely wild error case we expect to

dataplane init system part 1 #897

Are you sure you want to change the base?

dataplane init system part 1 #897

Uh oh!

Conversation

daniel-noland commented Oct 9, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Key changes:

Reviewed Changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Uh oh!

Copilot AI Oct 10, 2025

Choose a reason for hiding this comment

Uh oh!

daniel-noland Oct 10, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

qmonnet left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

qmonnet Oct 10, 2025

Choose a reason for hiding this comment

Uh oh!

mvachhar Oct 10, 2025

Choose a reason for hiding this comment

Uh oh!

qmonnet Oct 10, 2025

Choose a reason for hiding this comment

Uh oh!

qmonnet Oct 10, 2025

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

daniel-noland commented Oct 9, 2025 •

edited

Loading

qmonnet left a comment •

edited

Loading