Re: [GIT PULL] percpu fixes for 2.6.32-rc6

[Ingo Molnar]

* Tejun Heo <tj@xxxxxxxxxx> wrote:

> Hello, Ingo.
> 
> Ingo Molnar wrote:
> >> I was mostly worrying about introducing unrelated bug while changing. 
> >> Anyways, one patch it is.  I'll route it through tip as suggested by 
> >> Ingo in a few hours.
> > 
> > Btw., i'd suggest you keep it in your percpu tree as usual and send it 
> > to Linus directly - i offered testing for the cleanup patch (and can 
> > pull your patch for such a purpose), it doesnt 'have' to go via -tip.
> 
> Can you please then pull from the following tree for testing?  I'll 
> push it to Linus after a couple of days if nothing explodes.
> 
>   git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu.git for-linus
> 
> Thanks.

Sure - pulled it into tip:master for testing earlier today and after a 
few hours of it's looking good so far in x86 runtime tests. I also did 
cross-build testing to a dozen non-x86 architectures and it was fine 
there too.

btw., there's some 80-cols checkpatch warning artifacts in the commit:

+                       if (pcpu_extend_area_map(chunk, new_alloc) < 0) {
+                               err = "failed to extend area map of "
+                                       "reserved chunk";

which suggest that the logic here is perhaps nested a bit too deep. It 
could be improved by moving the reserved allocation branch of 
pcpu_alloc():

        if (reserved && pcpu_reserved_chunk) {

into a helper inline function, something like __pcpu_alloc_reserved().

It's a rare special case anyway. It could be changed to return with the 
pcpu_lock always taken, so the above branch would look like this:

	if (unlikely(reserved)) {
		off = __pcpu_alloc_reserved(&chunk, size, align, &err);
		if (off < 0)
			goto fail_unlock;
		goto area_found;
	}

Which is a cleaner flow IMO, and which simplifes pcpu_alloc().

And the error string should be:

		err = "failed to extend area map of reserved chunk";

(Ignore the checkpatch complaint.)

What do you think?

	Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/