Re: [PATCH RFC tip/core/rcu 32/41] rcu: Update stall-warning documentation

[Josh Triplett <josh@joshtriplett.org>]

On Thu, Feb 02, 2012 at 10:18:05AM -0800, Paul E. McKenney wrote:
> On Wed, Feb 01, 2012 at 09:56:39PM -0800, Josh Triplett wrote:
> > On Wed, Feb 01, 2012 at 11:41:50AM -0800, Paul E. McKenney wrote:
> > > From: "Paul E. McKenney" <paul.mckenney@linaro.org>
> > > 
> > > Add documentation of CONFIG_RCU_CPU_STALL_VERBOSE, CONFIG_RCU_CPU_STALL_INFO,
> > > and RCU_STALL_DELAY_DELTA.  Describe multiple stall-warning messages from
> > > a single stall, and the timing of the subsequent messages.  Add headings.
> > 
> > Don't some of these documentation changes go with earlier patches in
> > this series?
> 
> Some could, but there is a fair amount of catch-up here.  Since we don't
> need documentation to be bisectable, it makes sense to do a single
> commit to update the documentation.

I think we've actually had this particular conversation once for each of
the last few rounds of patches. :)

I agree that documentation doesn't have to allow bisection, but I do
think it generally makes sense to add documentation together with
whatever change it documents whenever possible.  Among other things,
doing so makes a series of patches much easier to rearrange and merge.

When documenting things that previously had no documentation, and didn't
appear in the same patch series, it makes sense to have a separate
commit to add documentation.  I'd just suggest that when documenting
things added or changed in the same patch series, the documentation
should go with the addition or change.

> > Also, this commit message doesn't say anything about the removal of
> > RCU_SECONDS_TILL_STALL_RECHECK:
> > 
> > > --- a/Documentation/RCU/stallwarn.txt
> > > +++ b/Documentation/RCU/stallwarn.txt
> > > @@ -14,12 +14,36 @@ CONFIG_RCU_CPU_STALL_TIMEOUT
> > >  	issues an RCU CPU stall warning.  This time period is normally
> > >  	ten seconds.
> > >  
> > > -RCU_SECONDS_TILL_STALL_RECHECK
> > [...]
> > > -	This macro defines the period of time that RCU will wait after
> > > -	issuing a stall warning until it issues another stall warning
> > > -	for the same stall.  This time period is normally set to three
> > > -	times the check interval plus thirty seconds.
> 
> It is now computed from CONFIG_RCU_CPU_STALL_TIMEOUT, which has an old
> value for default, which is now fixed.  I will add the rationale for
> removing CONFIG_RCU_SECONDS_TILL_STALL_RECHECK to the commit message.

Thanks.

- Josh Triplett