Re: [PATCH] sched: Support current clocksource handling in

[john stultz <johnstul@us.ibm.com>]

On Wed, 2009-05-27 at 09:26 +0900, Paul Mundt wrote:
> On Tue, May 26, 2009 at 05:22:10PM -0700, john stultz wrote:
> > On Wed, 2009-05-27 at 08:44 +0900, Paul Mundt wrote:
> > > What would you recommend instead? We do not want to spin here, and if we
> > > are in the middle of changing clocksources and returning jiffies anyways,
> > > then this same issue pops up in the current sched_clock() implementation
> > > regardless of whether we are testing for lock contention or not.
> > > Likewise, even if we were to spin, the same situation exists if the new
> > > clocksource does not have the _SCHED_CLOCK bit set and we have to fall
> > > back on jiffies anyways, doesn't it?
> > > 
> > > Put another way, and unless I'm missing something obvious, if we ignore
> > > my changes to sched_clock(), how is your concern not applicable to case
> > > where we are changing clocksources and using generic sched_clock() as it
> > > is today?
> > 
> > Well, Thomas' point that locking isn't necessary, as sched_clock()
> > doesn't have to be correct, is probably right. 
> > 
> > So, I think a get_sched_clocksource() interface would be ideal (if we
> > want to get academic at a later date, the pointer could be atomically
> > updated, and we'd keep it valid for some time via an rcu like method).
> > 
> > Additionally, you can set the jiffies clocksource as a _SCHED_CLOCK
> > clocksource and drop the jiffies fallback code completely.
> > 
> I thought about that initially as well, but in the case of the jiffies
> clocksource, that won't handle INITIAL_JIFFIES, which we want to subtract
> to make printk times sane.

Fair point, but that shouldn't be a big issue, we can fix that in the
jiffies clocksource read() implementation.

thanks
-john