OK. I started playing with this and I wrote a small kernel thread which is
awakened periodically by the interrupt service routine. Basically I have something
like this.

device_monitor()
{
    while(1)
    {
        tsleep(&do_monitor, PWCHAN, "bharani", 0); /* PWCHAN = 12 */
        Do_read_device_registers();
    }
}

and I have my interrupt service routine coded like this :

service_interr()
{
    ...
    ...
    wakeup(&do_monitor);
}

(I am running NetBSD on our hardware simulator and I put break points on the
kernel thread routine)


I can see the thread being created (when I do ps, I see "bharani"). But it does
not seem to be getting scheduled at all. Obviously I am missing something. I think
I am giving it a wrong priority or something. Could you please figure out what I
am doing wrong??

Thank you,
Bharani.

Bill Sommerfeld wrote:

> > When you say, extremely early boot, it is device configuration??
>
> Yes (well, at least the parts of device config which run before
> interrupts are enabled and the process scheduler is started).
>
> On some platforms (i386 comes to mind) there are dependancies which
> make it difficult to correctly initialize thread state before certain
> key drivers have been attached.
>
> > And I assume that the kthread_run_deferred_queue() is the one which calls all
> > these call-back functions. Am I right?
>
> Yes.