Subject: Re: "fast" syscalls?
To: Jason Thorpe <>
From: Garrett D'Amore <>
List: tech-kern
Date: 03/17/2006 13:46:31
Jason Thorpe wrote:
> On Mar 17, 2006, at 1:02 PM, Garrett D'Amore wrote:
>> No.  Application starts thread, which issues ioctl that never returns.
>> The ioctl code causes the thread to call tsleep when no data in the
>> queue.
>> Then I can "kick" that thread after adding stuff to the queue by calling
>> _lwp_wakeup().
> ...which is still a full-expense system call.  I would like to have
> the details of this hidden in some userland API, too... in particular,
> I don't think the pthread -> LWP binding is generally public data.

That was a concern of mine as well.  And I agree that the implementation
details should be locked away in a library.

For now I might just do heavy weight system calls and see how it performs.

Garrett D'Amore, Principal Software Engineer
Tadpole Computer / Computing Technologies Division,
General Dynamics C4 Systems
Phone: 951 325-2134  Fax: 951 325-2191