A clockevent IRQ delivery question

[Jon Loeliger]

On Fri, Jun 20, 2014 at 10:31 AM, Jon Loeliger <loeliger at gmail.com> wrote:

> OK, I'll leave things as they stand in my port for now
> as that appears to be working even if not ideal yet!

"Appears" is the key word here.  Sometimes it boots fully to a prompt,
and sometimes it doesn't, mostly depending on which printk()s I leave in
or take out.  So something flakey or timing sensitive perhaps.

If I manually inspect the timers from a different CPU in the system,
I can see them counting and reloading, but I am not getting
an interrupt delivered to some CPU.  I am trying to determine if it
is a broken timer IP block that is failing to generate an interrupt, or if
is a failed IRQ delivery to one of the cores.  Both cores are in the
idle loop.  If I manually trigger an interrupt, the kernel panics and
claims that a CPU was non-responsive for something like 50 seconds.

Do I need to do some form of IRQ broadcast to ensure all CPUs are
receiving the timer ticks?  How do I set that up?  Pointer to an example?

Thanks,
jdl