re: [patch] sys_epoll ...

[Davide Libenzi]

On Mon, 21 Oct 2002, Dan Kegel wrote:

> Davide wrote:
>  >asmlinkage int sys_epoll_create(int maxfds);
>  >asmlinkage int sys_epoll_ctl(int epfd, int op, int fd, unsigned int events);
>  >asmlinkage int sys_epoll_wait(int epfd, struct pollfd **events, int timeout);
>
> Hey Davide,
> I've always been a bit bothered by the need to specify maxfds in
> advance.  What's the preferred way to handle the situation
> where you guess wrong on the value of maxfds?  Create a new
> epoll and register all the old fds with it?  (Sounds like
> a good job for a userspace wrapper library.)

I'm currently looking at doing it automatically, w/out the need of
specifying it. I don't like it either ...

> Regardless, thanks for pushing /dev/epoll along towards inclusion
> in 2.5.  I'm looking forward to seeing it it integrated.
> Even if the interface doesn't please everyone, the performance
> should...

You should thank Hanna Linder that bugged Linus way more than I did :)

- Davide