Re: [PATCH 08/16] dma-debug: add core checking functions

[Joerg Roedel]

On Sun, Jan 11, 2009 at 08:57:52AM +0100, Joerg Roedel wrote:
> On Sun, Jan 11, 2009 at 12:11:27AM +0100, Ingo Molnar wrote:
> > 
> > * Joerg Roedel <joerg.roedel@xxxxxxx> wrote:
> > 
> > > +#define err_printk(dev, format, arg...) do {			\
> > > +		error_count += 1;				\
> > > +		if (show_all_errors || show_num_errors > 0) {	\
> > > +			WARN(1, "%s %s: " format,		\
> > > +			     dev_driver_string(dev),		\
> > > +			     dev_name(dev) , ## arg);		\
> > > +		}						\
> > > +		if (!show_all_errors && show_num_errors > 0)	\
> > > +			show_num_errors -= 1;			\
> > 
> > Note that the arithmetics here is SMP-unsafe: we only hold the hash bucket 
> > so if two errors hit at once on two CPUs then the error tracking variables 
> > can be accessed at once.
> > 
> > I'd suggest a simple global lock for this error case (taken inside the 
> > hash bucket lock), to be on the safe side.
> > 
> > Also, please dont use a macro for this - printk details can be passed in 
> > to helper inlines/functions too.
> 
> Yeah, this is not SMP-safe, I know. But debugfs does not support
> atomic_t so I made the variables u32. But at least a race condition has
> not a too bad impact. What may habben is that error_count misses a error
> or the show_num_errors become negative.
> But if we really want to avoid this I think its better to add atomic_t
> support to debugfs. What do you think?

Even a global lock will not really help here because show_num_errors and
show_all_errors can be set using debugfs. Either we live with the small
race (with limited impact) or I add atomic_t support to debugfs.

Joerg

--
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/