lkml.org 
[lkml]   [2012]   [Mar]   [28]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 11/32] nohz/cpuset: Don't turn off the tick if rcu needs it
On Tue, Mar 27, 2012 at 11:13:39AM -0500, Christoph Lameter wrote:
> On Tue, 27 Mar 2012, Frederic Weisbecker wrote:
>
> > > Is there any way for userspace to know that the tick is not off yet due to
> > > this? It would make sense for us to have busy loop in user space that
> > > waits until the OS has completed all processing if that avoids future
> > > latencies for the application.
> >
> > What is the usecase you have in mind? Is it for realtime purpose?
>
> Please do not use "realtime" since I am not sure what you mean by that.
> Its for a low latency applications that cannot use "realtime" because that
> implies high latencies.
>
> > The "tick stopped" is a volatile and relative state.
>
> The use case is an application that cannot tolerate the latencies
> introduced by timer tick processing. It will only start running when the
> system is in a sufficiently quiet state.
>
> > Relative because if a timer list is enqueued to fire 1 second later,
> > the tick will be stopped until that happens. How do we consider this (common)
> > case?
> >
> > Also as Chris noted it is volatile because the tick can be restarted anytime
> > for random reasons: the CPU receives an IPI which makes it restart the
> > periodic tick.
>
> Ok some sort of notification would be good for that case. If a timer tick
> happens and that was unavoidable then it would be good to log the reason
> why this occured so that the system can be configured in such a way that
> these interruptions are minimized.

tracing is probably the right place to log these things. But that's
about debugging. This won't be a notification on top of which your app
could recover.


\
 
 \ /
  Last update: 2012-03-28 13:55    [from the cache]
©2003-2011 Jasper Spaans