CPU_LWP_FORK(9)

💅 CPU_LWP_FORK(9) — the kernel dev girlies' manual

okay so bestie, welcome to the Kernel Developer's Manual page for CPU_LWP_FORK(9) — yes that's the real section number, we're in the big leagues now 🎀

---

✨ NAME ✨

cpu_lwp_fork, child_return, lwp_trampoline — finish a fork operation. like, the actual finishing touches, not just the vibe 🔥

---

🌸 SYNOPSIS 🌸

#include <sys/proc.h>

void
cpu_lwp_fork(struct lwp *l1, struct lwp *l2, void *stack,
    size_t stacksize, void (*func)(void *), void *arg);

void
child_return(void *arg);

copy that #include exactly, bestie, no cap 🫶

---

🦋 DESCRIPTION 🦋

Okay so cpu_lwp_fork() is the machine-dependent portion of fork1() that finishes a fork operation, with the child lwp l2 nearly set up — like, it's at 99% and we just need to hit send 💅 It copies and updates the PCB and the trap frame from the parent l1, making the child ready to run. literally that's the whole glow-up.

cpu_lwp_fork() rigs the child's kernel stack so that it will start in lwp_trampoline(). and lwp_trampoline() does not have a normal calling sequence — it's entered by cpu_switchto(), so don't come at it expecting a regular call, it's built different 😭

If an alternate user-level stack is requested — and that means non-zero values in both the stack and stacksize arguments, both of them, no half-sending — the user stack pointer is set up accordingly.

After being entered by cpu_switchto() and while running in user context (within the kernel), lwp_trampoline() will invoke the function func with the argument arg. iconic behavior honestly ✨

• If a kernel thread is being created, the return path and argument are specified with func and arg.
• If a user process is being created, fork1() will pass child_return() and l2 to cpu_lwp_fork() as func and arg respectively.

This causes the newly-created child process to go directly to user level with an apparent return value of 0 from fork(2), while the parent process returns normally. so the child's like "hiii i returned 0" and the parent's like "i returned your pid" — both slaying, different paths 👑

---

🌈 SEE ALSO 🌈

• fork(2)
• cpu_switchto(9)
• fork1(9)

read them, they're the supporting cast and they ate 🍓

---

ngl this page is short but it's dense — every pointer matters. don't skip the "both non-zero" bit on stack and stacksize, that's the trap 🙈

---

and that's the page, bestie!! this one came from Debian, dated April 26, 2018 — vintage but still serving 💖 k byeee, go fork something responsibly 🧚‍♀️⭐