On Thu, Oct 25, 2007 at 05:40:03PM +0200, Jan Engelhardt wrote:
> On Oct 21 2007, Alexey Dobriyan wrote:
> >
> >One of type of bugs steadily being fought is the following one:
> >
> >	unsigned int flags;
> >	spin_lock_irqsave(&lock, flags);
> >
> >where "flags" should be "unsigned long". Here is far from complete list 
> >of commits fixing such bugs:
> >
> 
> How about making spin_lock_irqsave actually take a pointer to flags? 

How would you do it without flag day?

> (Which would be the logical choice if it were a function and not a 
> macro...) That would flag up all violations ("without cast to different 
> pointer" or so) while usually not breaking compilation.
> 
> Of course, irq_flags_t is probably the best long-term solution if one 
> wants to hide a struct. (Even then perhaps, use a pointer instead?)

IIRC, Christoph mentioned:

	irq_flags_t flags;

	flags = spin_lock_irqXXX(&lock);
	spin_unlock_irqYYY(&lock, flags);

where XXX and YYY are still to be found good names :^) It's also a solution
without flag day and with more sane lock part -- "how flags are modified
if they are passed by value?"

I start to like this proposal but I can't come up with good names.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/