I'm not a number, I'm a free file descriptor!!1
summary
Shows how Unix domain sockets carry open descriptors (use socketpair) to bypass per-process fd limits (Linux default ~1,000, OS 10 ~7,000) and how recursive passing creates 200,000 open files on Linux before the kernel chokes. Demonstrates a tiny test program that succeeds on Linux but fails on OS 10 because BSD closes trapped sockets, and reveals Linux’s unexpected mark-and-sweep garbage collector that traverses socket-contained fds about four–five levels deep.
AI-written summary · about
chapters (4)
related talks
01h 03mLinux Namespaces for Fun and Isolation
54mContainers unplugged: Linux namespaces
43mUnix Domain Socket: A Hidden Door Leading to Privilege Escalation in the Android Ecosystem
47mLinux user namespaces: a blessing and a curse
32mLinux user namespaces: a blessing and a curse
54mContainers unplugged: understanding user namespaces
54mUnderstanding user namespaces
15mPrivilege Separation and Pledge
