futex_cmpxchg(9)

📦 version 7.3.1, in nixpkgs unstable

💅 futex_cmpxchg(9) — the Kernel API Manual, bestie edition

Okay so picture this: it's a futex_cmpxchg(9) kind of day, and you're about to do an atomic compare and exchange. Yes, atomic. As in, no other thread can sneak in and ruin the vibe. Periodt. 💖

✨ SYNOPSIS ✨

u_int32_t futex_cmpxchg (futex_t *uaddr , u_int32_t oldval , u_int32_t
newval );

Look at her. Three args, one u_int32_t coming back at you. She's simple, she's iconic, she does exactly one thing and does it flawlessly. 🎀

🌸 ARGUMENTS 🌸

  • uaddr — the address of the futex you wanna modify. It's giving "main character address" energy, honestly. This is the spot. 🔥
  • oldval — the value you're expecting the futex to have. Like, if it isn't this, the whole thing doesn't go down. No cap.
  • newval — the new value you're tryna assign to that futex. Your glow-up value. ✨

🦋 DESCRIPTION 🦋

So basically this bad boy implements cmpxchg using the gcc atomic builtins. That's the whole thing. That's the assignment. 💅

If you want the lore, the original docs live here:

http://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html

And here's the part that matters: it returns the old futex value. Like, whatever was sitting there before you made your move. Lowkey obsessed with that — it lets you know if your exchange actually ate or if someone beat you to it. 🙈

🔥 SEE ALSO 🔥

The kernel file, bestie — go read it, iykyk:

Kernel file ./tools/testing/selftests/futex/include/futextest.h

And the whole futex friend group, all here, all slaying:

fr fr, this is the Kernel API Manual, dated October 2026, and the source line trails off at ./tools/testing/selftests/fute... — that's not me being mysterious, that's just literally where the footer cuts off. 🫶

Okay that's the whole page, bestie. Atomic compare and exchange, documented, understood, and honestly? She ate. 💅✨🎀🦋🌸💖🔥👑

📦 where this page comes from