Re: Real-Time Preemption and RCU

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

On Fri, Mar 18, 2005 at 06:11:26PM +0100, Ingo Molnar wrote:
> 
> * Paul E. McKenney <paulmck@us.ibm.com> wrote:
> 
> > For the patch, here are my questions:
> > 
> > o	What is the best way to select between classic RCU and this
> > 	scheme?
> > 
> > 	1.	Massive #ifdef across rcupdate.c
> > 
> > 	2.	Create an rcupdate_rt.c and browbeat the build system
> > 		into picking one or the other (no clue if this is
> > 		possible...)
> > 
> > 	3.	Create an rcupdate_rt.c and rely on the linker to pick
> > 		one or the other, with rcupdate.h generating different
> > 		external symbol names to make the choice.
> 
> you can also go for option #0: just replace the existing RCU code with
> the new one, and i'll then deal with the configuration details.

Having just spent the past few minutes descending into #ifdef hell,
I agree completely with your option #0.

> what will have to happen is most likely #2 (since there is near zero
> code sharing between the two variants, right?). Picking rcupdate_rt.c is
> as simple as doing this:
> 
>  obj-$(CONFIG_DONT_PREEMPT_RCU) += rcupdate.o
>  obj-$(CONFIG_PREEMPT_RCU) += rcupdate_rt.o
> 
> and then use Kconfig to generate either CONFIG_DONT_PREEMPT_RCU
> (default) or CONFIG_PREEMPT_RCU (if the user selects it).

Cool!  Thank you for the tutorial!  And yes, there is no shared code
that I can see.

> but it's not yet clear whether we want to offer this to users as a
> configurable option. The simplest solution for you would be to go with
> option #0 :-) [or if you prefer switchability, #1 is good too - i can
> then extract the bits and do #2 based on that.]

Option #0 it is -- I will stick with the locking algorithms and let
wiser and more experienced heads work out the configuration issues.

> > o	How best to interface to OOM?  Left to myself, I leave this
> > 	for later.  ;-)
> 
> yeah, i'd not worry about OOM that much at this stage.
> 
> > I will take the cowardly approach of patching against the upstream
> > kernel.
> 
> sure. This is in fact easier for me: i'll first rip all my RCU hackery
> out of -RT and then add your code, so the base i'll be merging against
> will be closer to upstream than to current -RT.

Sounds good to me!

						Thanx, Paul