Re: [GIT RFC] percpu: use dynamic percpu allocator as the default percpu allocator

[Tejun Heo <tj@kernel.org>]

Rusty Russell wrote:
> On Thu, 9 Apr 2009 09:23:25 pm Tejun Heo wrote:
>> It's generally a good idea to use uniquely
>> distinguisible identifier for static symbols anyway to help debugging.
> 
> Sorry, I can't let this statement stand.  It's completely wrong: use the
> shortest clear name, always.

Heh... well, yeah, I don't think this is something everyone can agree
about.  I generally like having proper prefix for static symbols.  It
makes symbols more consistent too.

>> If this limitation is acceptable, I think we should also add the dup
>> build failure thing to the generic definition too tho so that such
>> cases can be discovered before they hit alpha and s390 later.
>>
>> Any objections?
> 
> Yes.  If we decide that static per-cpu is unsupportable, let's not
> hide the damn thing.  We just make it give a compile warning if we
> can, patch out the current cases, and make checkpatch.pl warn on new
> ones.
> 
> Don't silently override "static".  Don't come up with stupid
> justifications.  Accept with open-eyes that it's evil, just a lesser
> evil.

It's different tho.  With the right tricks, we can still make percpu
variables defined static unusable from other compile units.  ie. if we
make the combination of DECLARE_PER_CPU() and static DEFINE_PER_CPU()
trigger compile error, most of what the programmer meant by 'static'
can be achieved whether the symbol itself ends up visible outside of
the compile unit or not.

Thanks.

-- 
tejun