Linux Kernel & OS

Linux 7.5 introduces kcore= boot parameter to optionally restrict memory access

Key Takeaway

A new boot parameter allows administrators to disable /proc/kcore to prevent processes with CAP_SYS_RAWIO from reading the entire system memory.

Summary

A new patch queued for the Linux 7.5 merge window introduces a mechanism to optionally disable access to the /proc/kcore virtual file at boot time. This file, which provides a view of physical memory and kernel virtual address space in ELF format, is a significant security concern if a process gains high-level hardware access privileges.

While /proc/kcore is essential for live kernel debugging and memory forensics, it currently presents a large read primitive for any process possessing the CAP_SYS_RAWIO capability. The upcoming change allows system administrators to disable this interface using a new boot parameter without needing to compile custom kernels or use restrictive lockdown modes.

Security implications of /proc/kcore access

  • New Boot Parameter: The kcore=off parameter allows users to disable the /proc/kcore interface during the boot process, provided the kernel was compiled with CONFIG_PROC_KCORE.
  • Privilege Escalation Risk: Access to /proc/kcore is currently granted to any process with CAP_SYS_RAWIO, which can expose the entire kernel virtual address space and reveal KASLR (Kernel Address Space Layout Randomization) layouts.
  • Granular Control: Unlike the lockdown=confidentiality mode, which disables various debugging features like kprobes and BPF probes, this new parameter provides a surgical way to disable memory reading via procfs while keeping other kernel features intact.
  • Implementation Status: The patch has been integrated into the vfs-7.5.procfs branch of the VFS Git repository and is expected to enter the Linux 7.5 merge window soon.

Why it matters

This change simplifies security hardening for system administrators and cloud providers who need to run a single, standard kernel image across both debugging-heavy and high-security environments. It allows for better isolation by ensuring that hardware-level access (CAP_SYS_RAWIO) does not automatically grant full system memory visibility.

Original reference: www.phoronix.com →