AW: AW: Interrupt handling

Thanks Espen for the clarification of the interrupt processing! Finally
my handmade user mode echo server works. :-)

The problem I've been tracking for hours was this:
-> At the end of root tasks main I made it do some bogus calculation
forever: while ( true ) i = (i + 1) % 33331; (i is volatile, of course)

-> Even that I've started the echo server thread, it got never
scheduled. I guess a new thread has a lower priority than the root
thread by default, even if both are in the same address space. This is
why the interrupt got ACK and MASK, but never processed.

-> In the beginning I tried to send the root task sleeping forever:
   L4_Sleep( L4_Never );
   But if the echo_server is waiting for an interrupt IPC the kernel
sends a lot of "No runnable threads" (or something similar) messages,
which are annoying.

-> I solved the problem by creating a second thread, which does the
bogus calculation and sending the root task to sleep forever.

The wrong threadid would have been the next problem to encounter, but
fortunately I got the hint before. :-)

However, I found some strange behaviour of the kernel during my
debugging. In the beginning, I moved the interrupt registration into the
echo_server thread, so it would have been executed in the beginning. But
as explained, the thread got never scheduled. That is, the interrupt was
never registered. If I triggered an interrupt, I got the following
message:
<console>
Disable MPSC_INT_RCC
ACK MPSC_INT_RCC
Assertion irq_tcb->get_state().is_halted() failed in file
/cygdrive/c/Martin/pro
jects/L4-Pistachio/kernel/src/api/v4/interrupt.cc, line 169
(fn=c0025d40)
>> KD# assert
IP: 0xc002ddec, MSR: 0x00023032, user IP: 0x00300410, user MSR:
0x0000d032
</console>

That shouldn't happen, I assume. Is it expected, that all interrupts are
masked unless they are registered? Because on registering for an
interrupt, a call to intctrl_t:mask() is done from somewhere inside the
kernel.

Best,

Martin.