Re: [PATCH v3 4/4] sched_clock: Avoid deadlock during read from NMI

[Stephen Boyd <sboyd@codeaurora.org>]

On 02/04, Steven Rostedt wrote:
> On Wed, 4 Feb 2015 17:23:51 -0800
> Stephen Boyd <sboyd@codeaurora.org> wrote:
> 
> > On 01/30, Daniel Thompson wrote:
> > > @@ -98,26 +98,50 @@ unsigned long long notrace sched_clock(void)
> > >  }
> > >  
> > >  /*
> > > + * Updating the data required to read the clock.
> > > + *
> > > + * sched_clock will never observe mis-matched data even if called from
> > > + * an NMI. We do this by maintaining an odd/even copy of the data and
> > > + * steering sched_clock to one or the other using a sequence counter.
> > > + * In order to preserve the data cache profile of sched_clock as much
> > > + * as possible the system reverts back to the even copy when the update
> > > + * completes; the odd copy is used *only* during an update.
> > > + */
> > > +static void update_clock_read_data(struct clock_read_data *rd)
> > 
> > notrace?
> 
> Why? Isn't this for update, not for readers?
> 

Good point. I was basing it on the fact that the caller,
update_sched_clock() is marked notrace. It looks like it's always
been that way (see commit 2f0778afac79 "ARM: 7205/2: sched_clock:
allow sched_clock to be selected at runtime" from 2011-12-15
where it was introduced). So the correct thing would be to drop
notrace from update_sched_clock().

-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project