[LITMUS^RT] Is the hrtimer_start_on_info structure okay?

[John Gamboa]

Hello,

I'm trying to develop a scheduling plugin based on a table that describes
which task should be running in each core. My plan is to use hrtimers that
fire when a new task should be scheduled (a new entry in the table). I
wanted to start the timers in the "activate_plugin()" function, but there
is a problem: the hrtimer callback is executed normally by the same core
that calls the "hrtimer_start()" function (which starts the timer).

While reading the code, I've seem that Litmus introduces the
"hrtimer_start_on_info" structure in the hrtimer.h/.c files. If I
understood it right, they are exactly what I'm looking for. I planned to
use them to start the timer on other cpus, but after grepping for
"start_on_info" through the entire kernel code, I've seem that it is
virtually never used (an instance is declared inside the "rt_domain"
structure, but that is all).

Since my last question pointed to some left overs from old Litmus versions,
I though it would be better to ask: should I use them? Is there another
alternative I didn't notice? Is that structure another left over from the
past?


Thanks!
-- 
John Gamboa
rabanetescebolas.blogspot.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.litmus-rt.org/pipermail/litmus-dev/attachments/20121005/92f4c0be/attachment.html>