Linux 7.3-rc2 arrives with a broad set of testing-cycle fixes

Linux 7.3-rc2 was released on September 6 in UTC. Linus Torvalds's announcement email was sent at 23:26 UTC, already the early hours of September 7 in Baghdad. The Git record also confirms the September 6 release tag. This is the second release candidate, rather than the final 7.3 release.

Fixes across several areas

The announcement discusses filesystem, DRM, networking, BPF and driver fixes. Torvalds says tooling accounts for approximately 20% of the overall patch, particularly sched_ext and selftests, while drivers remain the dominant area. That describes the distribution of changes, not a percentage performance improvement for every computer.

What rc means

The official kernel development-process documentation describes a stabilization phase after the merge window, with rc releases generally appearing weekly to find and repair problems before the final version. This report does not assign a future completion date to 7.3.

For readers, the word “released” is not sufficient on its own. Knowing whether a version is a testing candidate or a stable release explains its role in development and avoids treating every version number as the same kind of milestone.

Give testing a defined question

In our assessment, a useful test needs a specific question. Does a particular device work correctly with a particular driver? Can a previously documented problem still be reproduced? Those questions produce more useful information than installing a new version and simply saying it appears to work.

A team can record the comparison version, configuration and steps needed to reproduce the issue. That makes the outcome understandable to another person, instead of an impression that cannot be reproduced. This is a suggested evaluation method; we have not installed the kernel or run performance benchmarks.

It also helps to identify the function being evaluated before starting. Successful booting alone does not answer a question about a different workflow. The test should match the original issue, and the conclusion should remain limited to what was actually examined.

Change volume is not a universal measure

A large batch of changes can contain several kinds of repair. Its size alone does not measure every device's speed or stability, nor establish an improvement for every application using the kernel.

For a developer, the value of this candidate lies in investigating an issue relevant to their work or the drivers they depend on. The precise reading of the announcement is a step within the testing phase, with fixes distributed across multiple areas. It is not an announcement that the cycle is finished or a universal performance result. Keeping those distinctions visible makes the release useful as development news without turning it into a claim it does not support.

Sources

• Linus Torvalds — September 6 release email, reproduced by LWN
• Linux mainline — v7.3-rc2 tag
• Linux 7.2 documentation — development process