Chasing down why installing the kernel segfaulted
The article details a debugging journey where a kernel installation segfaults, leaving the system unbootable. The author, running a rolling-release Linux distribution, encounters the error after updating the kernel package. Through systematic investigation, they isolate the cause to a race condition in the package manager's post-install script that generates the initramfs. The script, designed to run hooks concurrently, fails when two hooks attempt to modify the same temporary file simultaneously, resulting in a corrupted initramfs. The segfault occurs when the bootloader tries to load this corrupted image. The fix involves adding file locking or serializing the hook execution. The author notes that the bug is reproducible on systems with multiple CPU cores and high I/O load. The post includes detailed steps of using strace, dmesg, and manual initramfs regeneration to recover. The underlying issue is a missing synchronization primitive in the package manager's hook system, which has been reported upstream.
A race condition in package manager scripts can silently corrupt system boot.