Linux 7.3 Kernel Maintainers Overwhelmed by Flood of AI-Generated Patches

Kernel Maintainers Face Unprecedented AI-Generated Patch Volume

Linux 7.3 kernel maintainers are completely overwhelmed by AI as they process the latest networking subsystem updates during the Linux kernel merge window. Jakub Kicinski and co-maintainer Paolo Abeni merged 632 patches into the net tree and 648 into net-next, but revealed that between one-third to one-half of net-next patches appear to be AI-generated patches consisting of low-priority fixes or cleanups and clarifications1

1

. "We are completely overwhelmed, of course," Kicinski stated in the pull request, highlighting how AI and LLM agents have created relentless code churn that kernel development teams now struggle to manage1

1

.

The surge reflects a broader challenge facing open-source kernel maintainers: AI tools are generating patches faster than human reviewers can properly evaluate them. Many submissions target rarely-used drivers or address edge cases in APIs that have historically been racy, particularly for rare events like PCIe errors and timeouts1

1

. The networking developers have already dropped old networking code and drivers to reduce AI patch noise for components likely unused by anyone running modern, up-to-date kernels, yet the submissions continue1

1

.

Meta Provides LLM Resources to Combat AI Patch Overload

To address this crisis, kernel maintainers are fighting fire with fire. Meta has stepped in to provide sufficient LLM resources and access to help the team run reviews with multiple frontier models on each patch1

1

. This multi-model approach helps eliminate some hallucinations and improves patch review accuracy, though Kicinski acknowledges that LLMs can only do so much when it comes to substantive technical review1

1

.

The maintainers' strategy for the next release cycle involves shifting focus beyond just reviews. They plan to let LLMs take care of busy work—managing patchwork, automating common process complaints, editing commit messages, and potentially applying patches that already received "reviewed-by" tags from trusted contributors1

1

. Kicinski noted that certain complex issues, particularly race conditions in code paths like PCIe error handlers, cannot be resolved solely by AI and still require human expertise2

2

.

Linux 7.3 Delivers BIG TCP Support and New WiFi Drivers

Despite the patch management challenges, Linux 7.3 brings substantial networking subsystem updates. The headline feature is BIG TCP support for UDP tunnels, specifically the VXLAN and Geneve drivers1

1

2

2

. This enhancement allows the kernel's network stack to handle data chunks larger than 64 KiB, reducing per-packet overhead on high-speed connections2

2

. Previously exclusive to direct routing tools like Cilium, BIG TCP now extends to overlay networks commonly deployed in cloud and container environments2

2

.

Developer benchmarks show throughput improvements ranging from 10 to 15 percent under standard MTU conditions, with gains approaching 35 percent for VXLAN when hardware tunnel offload is disabled2

2

. The update also includes improved MPTCP handling of extreme memory pressure and enhanced error reporting for AF_UNIX1

1

2

2

.

New Hardware Support Expands Network Capabilities

Linux 7.3 introduces several new WiFi drivers including the mm81x driver for Morse Micro Long-Range S1G devices supporting Wi-Fi HaLow, and NXPWIFI for NXP devices1

1

2

2

. The Mediatek MT76 driver gained support for MT7928 and MT7925 NAN, while continued Multi-Link Operation enhancements improve wireless performance1

1

.

For wired networking, the update brings initial CXL support for AMD/Solarflare NICs, an initial skeleton Intel iXD driver for Intel E2100 and later IPUs and FNICs, and firmware flashing support for AMD Pensando NICs1

1

. The first release candidate for Linux 7.3 is expected around August 30, with stable release anticipated in late October2

2

. Watch how this AI-assisted review process evolves—it could set precedent for managing open-source contributions across major projects facing similar AI-generated submission floods.