Interrupt latency problems

[Paul Richards]

On Sat, 2003-05-10 at 19:02, Bosko Milekic wrote:

>   The thing is that in 5.x the Giant lock is more expensive in itself
>   and interrupts themselves are blocked on Giant.  Further, you have to


Ahh ok, I didn't know that the interrupts themselves were blocked in 5.x
which explains why some of them are just going missing. I was thinking
that at least the hardware interrupt handler got to run as it did in 4.x
but now that I think that through it doesn't make sense since the
handler would end up needing to lock on something pretty soon after
starting whereas in 4.x there'd only be one process in the kernel and
the interrupt could be serviced in the bottom half without getting in
the way of anything else.
 
-- 
Paul Richards