> Sven.Heursch@unibw-muenchen.de (Sven Heursch) writes:
>
> > "Stephen C. Tweedie" wrote:
> >
> > > Hi,
> > >
> > > On Wed, 14 Jul 1999 14:28:14 +0200, Sven Heursch
> > > <Sven.Heursch@unibw-muenchen.de> said:
> > >
> > > > Could it be that the scheduler in the kernel 2.2 has problems with the
> > > > priority based scheduling algorithm SCHED_FIFO and SCHED_RR. The
> > > > verification was made with two concurrent running processes with the
> > > > real time priorities 30 and 99. Normaly the higher one should be
> > > > scheduled first. But in Linux 2.2 that do not happend evereytime.
> > >
> > > What test case breaks the rule for you? The scheduler's goodness
> > > function should still absolutely prefer non-SCHED_OTHER processes.
> >
> > Yes, indeed, the kernel source code says so! But the reality is different.
> > For example:
> > We have two processes, A & B. A has the priority 99, B the priority 30. Both
> > are the only processes which are scheduled with SCHED_FIFO in the system.
> > Process A starts and goes in nanosleep(). Now B got the CPU and does for
> > example the access() system call. After a few micorseconds, an interrupt is
> > generated and interrupts the process B. Process A has installed an
> > interrupthandler for that interrupt before and is beeing signaled by the
> > ISR.
>
> You mean it called a kernel function to install an interrupt handler,
> and the handler does a send_sig() ?
yes, process A installs an interrupt handler and that one calls from inside the
kernel the kill_proc() syscall to send a signal.
>
>
> > Now the crazy thing: Wheather A has the highest priority in the system and
> > is now runnable, process B is being scheduled after the ISR in Kernel 2.2.
> > Process A is sleeping on.
>
> So it is never waken up, or just too late?
>
Process A recieves the signal and wakes up after process B has finished his job.
S.Heursch
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/