Re: [PATCH] fix bogus hotplug cpu warning

[Hugh Dickins <hugh@veritas.com>]

On Mon, 27 Aug 2007, Andrew Morton wrote:
> On Mon, 27 Aug 2007 21:37:14 +0100 (BST)
> Hugh Dickins <hugh@veritas.com> wrote:
> 
> > > So I agree with the patch, but not with its description.
> > 
> > I don't see which part of the description you disagree with, but please
> > do improve it if you can.
> 
> I'd change the description to
> 
> 
> current_is_keventd() is a stupid load of crap.  It does
> 
> int current_is_keventd(void)
> {
> 	struct cpu_workqueue_struct *cwq;
> 	int cpu = smp_processor_id();	/* preempt-safe: keventd is per-cpu */
> 
> but that comment (and the code itself) only make sense when
> current_is_keventd() is called by keventd.  In which case thers is no point
> in the function even existing!

I disagree.  The comment (your very own, I think?) and the code made
sense to me.  It's saying this is rather unusual, but safe against
preemption, because keventd is per-cpu (it might perhaps be better
if it said each keventd is bound to a cpu).

The code works in such a way that if it is a keventd, then it'll
correctly arrive at the answer yes; and if it's not a keventd,
then even if it's preempted to a different cpu in the middle,
the test won't match any keventd and so it'll correctly arrive
at the answer no.

But I'm content with the half-liner already there.

> 
> We need to use raw_smp_processor_id() so that non-keventd (or, more
> specifically, non-pinned-to-one-cpu) callers won't generate "using
> smp_processor_id() in preemptible" warnings.

Yes, that's true, raw_smp_processor_id() is needed to avoid the
warning; but it was already preempt-safe with smp_processor_id(),
just in danger of giving an inappropriate warning.

Hugh