Interesting detective work. I really wonder how in this modern era people still write in a language that's so easy to screw up. I started writing C in late 1984 but have zero desire to ever even see it again. Today I do Swift, but if I were writing an OS Rust makes way more sense that C (which I hear is coming in Linux). Debugging broken C code seemed to often came down to tripping through memory not yours because you miss counted something. In my commercial memory allocator in the 90's I had a lot of features to catch such over/under/mis-writes, but it seems better to have the language not even let you screw up in this way, rather than try to catch it in a runtime test.
For those who look at the comments first, the first paragraph to give you context.
"A few weeks ago, I found and reported CVE-2022-25636 - a heap out of bounds write in the Linux kernel. The bug is exploitable to achieve kernel code execution (via ROP), giving full local privilege escalation, container escape, whatever you want."
