Re: O_NONBLOCK on files

Benjamin LaHaise (bcrl@redhat.com)
Mon, 24 Sep 2001 17:30:29 -0400


On Mon, Sep 24, 2001 at 02:05:34PM -0700, Simon Kirby wrote:
> Yes, but this sucks. My whole intent was an interface design that would
> never need to context switch. I'm not sure if this is even possible, but
> if so it would be very nice from a userspace perspective.

Actually, this was one of the design concerns that I had in coming up with
the async io interfaces. The interface was designed to make use of the
upcoming syscalls-in-userspace stubs that x86-64 implemented and will be
passed on to x86 in 2.5. All completion events are passed through a ring
buffer with minimal locking on the kernel side which only needs thread
locking in userspace. This way the whole userspace to kernel transition
can be avoided by the server under sufficiently high load.

-ben
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/