[LAU] realtime problem on 2.6.25

[Joshua D. Boyd]

On Fri, 2008-04-25 at 11:23 -0400, Paul Davis wrote:

> Ingo's RT patches simply make the performance of a system using
> SCHED_FIFO-scheduled threads *better*. They have *absolutely nothing* to
> do with whether or not a normal user can successfully execute a program
> that wants to use SCHED_FIFO. that is determined by other non-kernel
> elements of your system (which parts varies by distro, but primarily PAM
> on most current distros).

If I may differ, it *sometimes* makes the performance of systems using
SCHED_FIFO-scheduled threads better.

When I tried it, my system fell to pieces, and I was unable to figure
out how to tune it back to working in a reasonable amount of time (say
40 hours).  So I've been sticking with SCHED_FIFO without Ingo's RT
patches.  At least this way things basically work with enough buffering.
Although somehow regular priority threads (and processes) still manage
to interfere with my SCHED_FIFO threads.

Maybe I should bring this up on the dev list again.