Re: [RFC PATCH 0/6] RCU: Preemptible-RCU

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Thu, Dec 13, 2007 at 09:38:04PM +0100, Ingo Molnar wrote:
> 
> * Gautham R Shenoy <ego@in.ibm.com> wrote:
> 
> > Hello everyone,
> > 
> > This patchset is an updated version of the preemptible RCU patchset 
> > that Paul McKenney had posted it in September earlier this year that 
> > can be found here --> http://lkml.org/lkml/2007/9/10/213
> > 
> > This patchset incorporates the review comments from Oleg Nesterov and 
> > Steven Rostedt.
> > 
> > The testing report of the patchset is as follows:
> > ====================================================================
> > Patch-stack:  	2.6.23-rc3 + cpu-hotplug patches from 
> > 		http://lkml.org/lkml/2007/11/15/239 + Preempt-RCU
> > 		patches.
> > Test:		RCU-Torture running parallelly with CPU-Hotplug
> > 		operations.
> > Duration:	24 hours.
> > Architectures:	x86,x86_64, ppc64.
> > ====================================================================
> > 
> > 
> > Currently it is based against the latest linux-2.6-sched-devel.git
> > 
> > Awaiting your feedback!
> 
> thanks Gautham, the patchset from you and Paul looks good to me and i've 
> applied it to sched-devel.git to get it tested and reviewed some more.
> 
> from the Nitpicking Police, there are a couple of minor style 
> problems/warnings with the code, you can see it via:
> 
>   scripts/checkpatch.pl --file kernel/rcu*.c
> 
> nothing serious - RCU is still one of the cleanest subsystems in the 
> kernel:
> 
>                                       errors   lines of code   errors/KLOC
> kernel/rcuclassic.c                        0             575             0
> kernel/rcupdate.c                          1             138           7.2

This one is the exception to checkpatch.pl's rule against "volatile".  ;-)

The volatile declaration is within the ACCESS_ONCE() macro that is used
within rcu_read_lock() and rcu_read_unlock() to force the compiler to
maintain ordering with respect to interrupt handler running only on that
same CPU.

> kernel/rcupreempt.c                        0             953             0
> kernel/rcupreempt_trace.c                  0             330             0
> kernel/rcutorture.c                        8             995           8.0

Hmmm...  Definitely some old whitespace issues here...

> the eventual goal would be to match:
> 
>    scripts/checkpatch.pl --file kernel/sched*.[ch]
> 
> output ;-)

We should be able to make some progress in that direction.  ;-)

							Thanx, Paul