Xen devel - Feb 2006 - Periodic timer interrupts in the Mini-OS (working?)

Hello!

I have a question about timer interrupts in the Mini-OS and
HYPERVISOR_set_timer_op. From printouts, it seems like the Mini-OS only
recieves timer events after calling block_domain:

void block_domain(u32 millisecs)
{
    struct timeval tv;
    gettimeofday(&tv);
    HYPERVISOR_set_timer_op(monotonic_clock() + 1000000LL * (s64) millisecs);
    HYPERVISOR_sched_op(SCHEDOP_block, 0);
}

i.e., not the "regular" 10ms periodic events (if I don''t call
block_domain(),
it will just sit idle). I''m trying to understand why this is the case.

>From looking in the Linux source code, I presume that the periodic ticker is
setup in setup_cpu0_timer_irq() on uniprocessors, and this simply does a
bind_virq_to_irqhandler(VIRQ_TIMER, ...), which is the same thing that the
mini-os does by calling bind_virq(VIRQ_TIMER, ...) in the init_time()
function. These two functions do the same things, including unmasking the
event channel.

The mini-os also does a __sti(), also taken from Linux, to enable event
delivery before init_time().


Is there anything else which needs to be done to deliver periodic timer
events?

-- 
// Simon

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 2006-02-02, 09:14, Simon Kagstrom wrote:
> Hello!
> 
> I have a question about timer interrupts in the Mini-OS and
> HYPERVISOR_set_timer_op. From printouts, it seems like the Mini-OS only
> recieves timer events after calling block_domain:
> 
> void block_domain(u32 millisecs)
> {
>     struct timeval tv;
>     gettimeofday(&tv);
>     HYPERVISOR_set_timer_op(monotonic_clock() + 1000000LL * (s64)
millisecs);
>     HYPERVISOR_sched_op(SCHEDOP_block, 0);
> }
> 
> i.e., not the "regular" 10ms periodic events (if I don''t
call block_domain(),
> it will just sit idle). I''m trying to understand why this is the
case.
It turned out it was my fault, for some reason I had modified do_event() to
include a mask_evtchn(port) in the end. So then, naturally, there were no more
events...

Sorry about the fuss, I should have checked the changes first.

-- 
// Simon

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel