Re: [PATCH] x86: remove byte locks

[Ingo Molnar <mingo@elte.hu>]

* Jiri Kosina <jkosina@suse.cz> wrote:

> On Tue, 13 Jan 2009, Jeremy Fitzhardinge wrote:
> 
> > > Why can't this just be somewhere in documentation? (possibly even with 
> > > the byte locks code as a reference).
> > Because Ingo's compil-o-matic will never fail on a documentation error.
> 
> Hmm, I have always considered the "we don't accept any code that would 
> have zero in-kernel users" rule as a quite reasonable one, at least in 
> order to prevent from bloat and code getting confusing.
> But apparently it's not the intention here.
> 
> > > It is IMHO just totally confusing to have a spinlock implementation that is
> > > not used at all in the tree. It took me quite some time to go through this
> > > until I finally figured out that this code is actually never used.
> > > Currently, on first sight it might seem that byte locks are used whenever
> > > CONFIG_PARAVIRT is set, which is not true.
> > Well, a comment next to the code explaining the rationale probably 
> > wouldn't go astray.
> 
> I still strongly feel that if the only purpose of the code in kernel is 
> "to provide example", then it belongs to documentation.
> 
> > > And apparently even Linus got confused by this, which also tells us
> > > something by itself, see [1].
> > > [1] http://marc.info/?l=linux-kernel&m=123144211719754&w=2
> > It tells us that Linus couldn't give a rat's arse about virtualization, 
> > which is just something we have to cope with ;)
> 
> I am afraid this has nothing to do with virtualization. It's simply 
> confusing when looking at the code.

i'd tend to agree, that area of code is quite complex already.

	Ingo