On Mar 16, 11:54am, Gregory Kritsch wrote:
> Third, interrupts.  I think my technique for adding an interrupt server
> for the modem was rather nasty (locore.s).  I noticed some neat looking
> code in the Zeus siop driver.  Has anyone given any thought to sharing out
> EXTER interrupts in a slightly nicer way?

  I had real fun with the EXTRN interupts for the Zeus.  It took me a
while to realize that processing the interrupt code at level 6
interrupted block I/O code that shouldn't be interrupted.  When I tried
to set splbio to level 6, it raised havoc with the clock timing.  I
finally settled on the current method of having the level 6 interrupt
handler get the Zeus SCSI status, save it, and schedule an sicallback to
actually process it.

> Finally, about spl*().  What should I do.  As the code stands, I have a
> conditional definition of spltty based on NCOM (the 16x50 driver is called
> com, since it is mostly based on the com driver from i386/isa).  What
> dependancies exist between the various spl levels (like the thing in
> if_sl.c).  Running with tty at 6 sounds like a bad idea offhand, what
> does everyone else think.  I don't really want to write a buffered driver
> like the internal serial port, but it looks like I may have to.

  Running with spltty at 6 isn't a very idea - you will probably find
that all timing based on the clock gets very messed up.

Michael

------------------------------------------------------------------------------