The Rust DMA dispute in the Linux kernel – An attempt at analysis

Contents

Linux founder and main developer Linus Torvalds has finally taken a stand on disputes that have been simmering for two weeks and made them clear: He will override the resistance to the DMA interfaces for Rust kernel code. According to him, maintainers of parts of the kernel who do not contribute to Rust for Linux simply have no say in how other parts of the kernel use this code.

Meanwhile, the whole thing has already caused a lot of damage, as two Linux developers have resigned in the course of it. In both cases, it is no longer about the Rust DMA blockade – but mainly about factors in the development and project management of the Linux kernel that have been repeatedly criticized for decades. The following text attempts to categorize some aspects of the situation, but first we will briefly outline the four most important factors in the current situation:

Blockade attitude with the Rust DMA code sets an avalanche in motion

Christoph Hellwig, a long-time contributor to much central Linux code, recently spoke out publicly against the inclusion of Rust code, which allows Rust drivers to use the C code he maintains for Direct Memory Access (DMA). Hellwig is bothered by having to deal with several programming languages in one project – and wants to prevent this "cancer" from penetrating into core areas of the kernel.

Asahi founder resigns

Among others, Hector Martin ("marcan"), who initiated Asahi Linux and the general efforts to run Linux-based operating systems on Apple systems with the company's own ARM processors, took offense at the statements. Among other things, Asahi Linux uses a kernel graphics driver written in Rust and utilizing DMA. This comes primarily from other developers. However, Martin has written lots of other kernel code for Apple M processors and systems built with them. He has also been able to incorporate parts of this into the main development branch of Linux – but for various reasons this has not yet been successful.

Martin not only criticized Hellwig's statements, but also pointed out, as he had done several times in previous months, how laborious it is to contribute code to Linux. He voiced this criticism on Linux mailing lists, but also pilloried the situation on Mastodon ("shaming on social media", as he himself described it). Two well-known kernel developers of the Linux graphics driver code publicly criticized him for this.

A short time later, Martin resigned as the maintainer of the kernel code for Apple systems with ARM-based CPUs. Shortly afterwards, his Mastodon account, which had many thousands of followers, was also no longer accessible. Less than a week later, he announced his resignation as lead developer of Asahi Linux in a lengthy blog post. At the same time, a blog post was published by the new project management, which consists of seven well-known Asahi developers. They have made it their top priority to simplify maintenance by incorporating many of the changes to the kernel required for Asahi into the main development branch of Linux. As a result, the team is putting the work on supporting new Apple computers with M3 and M4 processors, which has not yet been tackled anyway, even further on the back burner.

Another Linux developer retires

A short time later, Karol Herbst resigned from his position as co-supervisor of the Nouveau kernel driver. He emphasized that he had not been very active in the kernel community for some time in reviewing and supporting the driver. At the same time, he mentioned the final trigger for his withdrawal: the use of the term or symbol "the blue line" used by some ultra-right-wing groups in the USA within the Rust DMA discussion by Theodore Ts'o ("tytso"), a central and long-established kernel developer.

Torvalds remains silent for a long time

During the whole charade, Linus Torvalds initially only spoke publicly twice about the events: In replies to emails from Hector Martin about being pilloried. There, the Linux founder notes, among other things, that the Ashai founder himself may be the problem(1, 2).

However, as Hellwig published a few days ago, Torvalds said in private that he would override maintainer vetoes on Rust kernel code if necessary. On Friday morning, the Linux founder and main developer then made his own comments in a lengthy email. In it, he criticized Hellwig's attitude – and emphasized that maintainers who stay out of the Rust work have no say whatsoever in Rust code for the use of their C code.