Linux Security Summit Europe 2024

A retrospective on the work to restrict unprivileged user namespaces by default in Ubuntu 24.04. This presentation will cover the challenges, problems, and the solutions that Ubuntu choose. It will also take a look at work to address the problems that remain.

The Landlock security module lets Linux processes restrict what they can do and puts developers in charge of defining appropriate sandboxing policies for their programs. We will give a brief overview over Landlock’s current features, recent developments, and talk about what is next. We will discuss in more detail Landlock’s new support for restricting the use of IOCTL and the design considerations and trade-offs that went into it.

Restricting system calls can significantly reduce the attack surface. However, solutions like seccomp can be bypassed(CVE-2009-0835, CVE-2019-2054, CVE-2023-2431, etc.). If unused syscalls can be eliminated at config level and compile time, the attack surface can be fundamentally controlled.
However, the widespread presence of .pushsection in kernel code prevents linker to perform code garbage collection. The associated KEEP() directive also causes ownership reversal issues, resulting in related sections that should be removed to remain, leaving more unused code for potential exploitation by hackers.
By systematically reworking the .pushsection directive, we propose dead syscalls elimination. After specifying the syscalls that need to be retained, it can remove other syscalls' code without affecting the normal operation of the kernel. Attackers cannot exploit something that does not exist. This not only reduces the kernel size and eliminates the overhead of seccomp but also completely eradicates the possibility of exploitation.
Besides, the approach of eliminating the KEEP() directive can be generalized, further reducing the kernel's dead code and decreasing the attack surface.

Contributors: Tan Yuan, Fan Siqi, Liu Xiao, Wu Zhangjin, Liu Xin