Support io_uring

[yellowhatter]

Summary

This PR introduces io_uring support to Zenoh behind the uring feature flag.

The implementation is intentionally scoped as an internal transport/runtime enhancement rather than a public API change. The patch adds a new internal crate, commons/zenoh-uring, and wires it into the workspace so the io_uring-specific memory, batching, reader, and writer logic can live outside the public surface area.

What this changes

Transport integration

• Adds uring feature that is efficient only for Linux.
• Adds io_uring support to the Zenoh transport stack.
• Integrates the new I/O path with most links.
• Adds automatic read-mode selection between io_uring and tokio.
• Updates transport-side code paths to route io_uring operations through dedicated transport support code.

Link coverage

• Adds support for streamed/socket-based links.
• Keeps the existing transport model intact so the non-uring path still works normally.

Buffer and memory handling

• Introduces a dedicated page-backed arena for uring buffers.
• Allocates pinned memory with mmap/mlock and exposes it both as io_uring provided buffers and as registered iovec buffers.

Reader / writer path

• Adds dedicated reader and writer implementations for the uring path.
• Adds zero-copy oriented handling for split and fragmented payloads.
• Improves RX batch reclamation so large payloads can be handled with less memory waste.
• Tunes batch growth behavior to better support big messages and improve footprint.

Error handling and runtime integration

• Passes Tokio executor context into the uring thread so Tokio-aware user code remains compatible.

Build / CI / test coverage

• Adds the uring crate and dependency wiring to the workspace.
• Ensures Zenoh still compiles when the uring feature is disabled.
• Adds uring to CI.
• Expands uring-related tests, including dynamic-port coverage and extended UDP test paths.

Design notes

The implementation follows a conservative enhancement approach:

• no public API expansion,
• no behavioral change for users who do not enable uring,
• all uring-specific code isolated behind feature gates and internal crates,
• transport and buffer-management changes kept close to the existing architecture instead of introducing a separate parallel stack.

Expected user impact

For users building with -F uring, this adds a new Linux I/O backend that reduces overhead in the transport path and improve handling of larger or fragmented payloads. For users who do not enable the feature, the existing build and transport behavior remain unchanged. The improvement gives 25-50% latency reduction and few % messaging rate increase. Performance tests included in comments to this PR below.

Validation

Validation for this PR is centered on:

• successful compilation with and without uring,
• CI coverage for the new feature,
• transport-level tests for the new I/O path,
• UDP and split-buffer / fragmentation scenarios,
• correctness of buffer reclamation and RX finalization behavior.

---

🏷️ Label-Based Checklist

Based on the labels applied to this PR, please complete these additional requirements:

Labels: enhancement

✨ Enhancement Requirements

Since this PR enhances existing functionality:

• Enhancement scope documented - Clear description of what is being improved
• Minimum necessary code - Implementation is as simple as possible, doesn't overcomplicate the system
• Backwards compatible - Existing code/APIs still work unchanged
• No new APIs added - Only improving existing functionality
• Tests updated - Existing tests pass, new test cases added if needed
• Performance improvement measured - If applicable, before/after metrics provided
• Documentation updated - Existing docs updated to reflect improvements
• User impact documented - How users benefit from this enhancement

Remember: Enhancements should not introduce new APIs or breaking changes.

Instructions:

1. Check off items as you complete them (change - [ ] to - [x])
2. The PR checklist CI will verify these are completed

This checklist updates automatically when labels change, but preserves your checked boxes.

[codecov]

❌ 1 Tests Failed:

Tests completed	Failed	Passed	Skipped
835	1	834	0

View the top 1 failed test(s) by shortest run time

zenoh-uring::reader::linux_tests::rw_parallel_and_start_stop::tcp_reader_first_exclusive_reader

Stack Traces | 2.85s run time

2 errors occurred executing test:
* process aborted with signal 6 (SIGABRT)
* thread '<unnamed>' (15276) panicked at .../core/src/panicking.rs:233:5:
  panic in a destructor during cleanup
  thread caused non-unwinding panic. aborting.

To view more test analytics, go to the Test Analytics Dashboard
_{📋 Got 3 mins? Take this short survey to help us improve Test Analytics.}

[yellowhatter]

100Hz:

main:

uring:

[yellowhatter]

1kHz:

main:

uring:

[yellowhatter]

max freq

main:

uring:

[yellowhatter]

Docs update: zenoh-rs/zenoh-web#127