Kernel maintainers are stepping up the removal of legacy hardware drivers from the Linux source tree, citing a persistent flood of low-quality patches and bug reports generated by AI and large language model (LLM) coding agents. The latest targets include the Moxa serial card driver and several other aging drivers that serve hardware rarely seen on modern systems.
The pattern has been building for some time. AI-assisted coding tools, designed to contribute patches at scale, frequently latch onto old drivers and submit automated fixes or reports for hardware that no active user appears to be running. While individual submissions may seem harmless, the cumulative effect has become a real drain on reviewer time. Maintainers are now weighing whether it is worth continuing to carry drivers whose primary function seems to be attracting machine-generated noise.
The Moxa driver is a case in point. The serial card manufacturer's hardware is long discontinued from mainstream use, yet the driver remained in the tree as a matter of backward compatibility—a longstanding kernel tradition. With automated agents now peppering maintainers with context-free patches for such code, retirement has started to look like the more practical option.
This is not an isolated incident. The kernel has seen a broader cleanup wave targeting similarly obsolete drivers, and maintainers have signaled that more removals are likely to follow. The calculus is straightforward: when a driver supports hardware with no observable active user base and becomes a magnet for automated noise, the maintenance cost outweighs the benefit of keeping it around.
The trend raises practical questions for operators running legacy or embedded Linux systems. Drivers removed from the mainline kernel will eventually drop out of downstream distributions, potentially breaking systems that depend on niche hardware support. Users in industrial or specialized environments who rely on older peripherals may find themselves scrambling for alternatives once upstream support disappears.
For developers building AI coding tools, the situation highlights a significant blind spot. Current LLM agents can generate syntactically correct patches but lack the judgment to determine whether a given driver is still relevant, whether the hardware it supports is still manufactured, or whether the kernel community considers it worth maintaining. Until these tools develop that kind of contextual awareness, human maintainers will bear the burden of sorting useful contributions from noise.
The Linux kernel's handling of its aging driver portfolio is becoming a test case for how large open-source projects manage the intersection of legacy support and automated development. As AI-generated contributions continue to scale, the question of when to stop maintaining old code—and how to communicate that decision to affected users—is one that extends well beyond a single kernel subsystem.
核心維護者正加強從Linux源代碼樹中移除傳統硬件驅動程式，理由是AI和大型語言模型（LLM）編碼代理持續產生大量低質量補丁和錯誤報告。最新目標包括Moxa序列卡驅動程式及其他數個較老舊的驅動程式，這些驅動程式所服務的硬件在現代系統上已鮮見踪影。
這種模式已醞釀一段時間。旨在大規模貢獻補丁的AI輔助編碼工具，經常鎖定舊驅動程式，並針對似乎已無活躍用戶運行的硬件提交自動化修復或報告。雖然單項提交看似無害，但其累積效應已對審核時間造成實質消耗。維護者現正權衡是否值得繼續保留那些主要功能似乎只是吸引機器生成噪音的驅動程式。
Moxa驅動程式便是一個典型案例。這家序列卡製造商的硬件早已停止主流使用，然而該驅動程式因向後兼容性——一項悠久的核心傳統——而保留在代碼樹中。隨著自動化代理現在為這類代碼大量發送缺乏上下文信息的補丁，淘汰已開始成為更實際的選項。
這並非孤立事件。核心已見一波更廣泛的清理浪潮，目標類似過時的驅動程式，維護者亦表示可能會有更多移除行動。考量很直接：當一個驅動程式所支持的硬件無可觀察的活躍用戶群，並成為自動化噪音的吸引點時，其維護成本便超過了保留它的益處。
此趨勢為運行傳統或嵌入式Linux系統的營運者提出了實際問題。從主流核心移除的驅動程式最終將從下游發行版中消失，可能破壞依賴小眾硬件支持的系統。工業或專業環境中依賴舊周邊設備的用戶，可能在頂層支持消失後，不得不倉促尋找替代方案。
對於開發AI編碼工具的人員而言，這種情況凸顯了一個重大盲點。現時的LLM代理能生成語法正確的補丁，但缺乏判斷力來確定特定驅動程式是否仍具相關性、其支持的硬件是否仍在生產，或核心社區是否認為它值得維護。在這些工具發展出那種上下文感知能力之前，人類維護者仍需承擔從噪音中篩選有用貢獻的負擔。
Linux核心處理其老舊驅動程式組合的方式，正成為大型開源項目如何管理傳統支持與自動化開發交集的一個測試案例。隨著AI生成的貢獻持續擴大，何時停止維護舊代碼——以及如何將此決定傳達給受影響用戶——的問題，已遠超單一核心子系統的範疇。