Re: [PATCH] bug: Use xchg() to update WARN_ON_ONCE() staticvariable

[Steven Rostedt]

On Tue, 15 Oct 2013 13:12:59 -0700
Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:

 
> Also, we're now incurring an atomic op for every "call".  Presumably
> these calls are rare, but not necessarily - one can envisage uses of a
> generic ONCE() which are called at high frequency.  Should we avoid
> that with
> 
> #define ONCE() ({
> 	static int state;
> 	int ret;
> 
> 	if (likely(state))
> 		ret = 0;
> 	else
> 		ret = !xchg(&state, 1);
> 	ret;
> })

I was talking with Boris on IRC about having a shortcut if "state" is
already true. I argued against it, but that was just for the WARN()
operations because I rather add a LOCK xchg, then more branches to hot
paths.

But for a generic ONCE() function, I guess we would want the shortcut
as it may be used by something that gets hit all the time with a normal
kernel.

-- Steve
--
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/