Re: acpi_evaluate_integer broken by design

[Pavel Machek]

Hi!

> > Subject: acpi_evaluate_integer broken by design
> 
> Please don't give patches daft and not very meaningful titles.  It just
> means that someone else has to invent a useful title and then we get
> the same patch floating about with two different titles.

Sorry. I was kind of angry at that code because it does not contain
any remarks when it is okay to call it, etc.

> > Date: 	Tue, 25 Nov 2008 12:05:08 +0100
> > Sender: linux-kernel-owner@xxxxxxxxxxxxxxx
> > User-Agent: Mutt/1.5.18 (2008-05-17)
> >
> > 
> > Now I know why I had strange "scheduling in atomic" problems:
> > acpi_evaluate_integer() does malloc(..., irqs_disabled() ? GFP_ATOMIC
> > : GFP_KERNEL)... which is (of course) broken.
> 
> That is kinda weird.  When did this all start happening?

With the HP LED driver... so this is "would be nice for 2.6.28, not
worth backporting" material.

> > There's no way to reliably tell if we need GFP_ATOMIC or not from
> > code, this one for example fails to detect spinlocks held.
> > 
> > Fortunately, allocation seems small enough to be done on stack.
> 
> It's 24 bytes in an x86_64 allmodconfig build.  Clearly it should be a
> local.

Yep.
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
--
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/