A security researcher has documented a dramatic difference in how quickly vulnerabilities are addressed in the experimental Ginux Rust kernel versus the upstream Linux kernel. The project Ginux, initiated by Gregory Kurtzer (co-founder of CentOS and creator of Rocky Linux), aims to build an x86-64 UEFI-compatible kernel in Rust with Linux userspace compatibility, relying heavily on AI agents for development.
The researcher ran the Ginux codebase through automated analysis tools and conducted three independent vulnerability searches using multiple large language models. Fifteen issues were reported through GitHub, including a critical flaw in the TLS client for container image registries. The RNG was seeded only with RTC time at boot with one-second precision, causing predictable client_random values and ephemeral keys that allowed passive decryption of recorded TLS traffic.
Responses from maintainer agents arrived within four hours, and the first three fixes were merged into main within 10 hours. The TLS RNG issue was resolved approximately 12 hours after reporting. By the morning of November 2, five fixes had landed in main while ten additional PRs awaited review.
In contrast, the same researcher previously submitted 16 findings across three Linux subsystems, including Android Binder, net/xfrm IP-TFS, and nova-core. One patch for the Rust Binder driver, which replaced KVec with KVVec to avoid unnecessary physically contiguous memory allocations, took 69 days to reach Linus Torvalds' tree after initial submission on August 17. The four-line change addressed a condition where roughly 900,000 file descriptors could trigger WARN messages due to excessive contiguous memory requests.
The researcher noted that Linux follows rigorous compatibility and regression-avoidance processes across billions of devices, while Ginux remains an explicitly experimental prototype. The timeline comparison shows the first Linux response arrived within a day via email, whereas Ginux agents responded in roughly four hours.


