Re: Kqueues and fork
- In reply to: Konstantin Belousov : "Re: Kqueues and fork"
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Fri, 29 Aug 2025 14:55:36 UTC
On Fri, Aug 29, 2025 at 5:27 PM Konstantin Belousov <kostikbel@gmail.com> wrote: > > Then why don't you just open another kqueue in the child, for the > kqueue-based implementation? If wanted, you can dup2() it into the > same fd as the parent kqueue fd. If I understand it correctly, it'd require me to either 1. In the every pollable set method first check if kqfd is actually a live kqueue fd. That'd mean an extra fcntl(fd, F_KINFO) on each call 2. Add a new pollable_set_rearm_after_fork method and put a call to it after all forks in the code base. Both of these solutions seem less optimal than KQUEUE_CPONFORK, which allows pollable-set-kqueue to be a self-contained implementation.