Re: sysfs file creation result nightmare

[Jean Delvare <khali@linux-fr.org>]

Ben,

On Sun, 10 Dec 2006 11:55:31 +1100, Benjamin Herrenschmidt wrote:
> > And (ultimately) make the function return void.
> > 
> > Yes, that's probably a valid approach - we've discussed it before but nobody has
> > taken it further.
> 
> I would have preferred that approach (with a WARN_ON rather than a BUG
> though). On the other hand that would make it slightly harder for the
> few cases (if any ?) who actually want something like a "create if it
> doesn't exist already" semantic.

Let's just boldly state that nobody wants that semantic, if it helps.

> I'm a bit worried by the amount of code added by systematic checking of
> the results for cases that really should never happen. That's why I
> prefer a BUG/WARN type semantic.
> 
> Maybe the best is to have the examples like radeonfb actually do the
> 
> WARN_ON(sysfs_create_file(...));

Beware that sysfs_remove_bin_file() will complain loudly if you later
attempt to delete that file that was never created.

-- 
Jean Delvare