Re: [PATCH] EDAC: core EDAC support code

[Greg KH <greg@kroah.com>]

On Mon, Mar 06, 2006 at 10:14:22AM -0800, Dave Peterson wrote:
> On Sunday 05 March 2006 02:30, Arjan van de Ven wrote:
> > On Sun, 2006-03-05 at 11:18 +0100, Arjan van de Ven wrote:
> > > > +/* Main MC kobject release() function */
> > > > +static void edac_memctrl_master_release(struct kobject *kobj)
> > > > +{
> > > > +	debugf1("EDAC MC: " __FILE__ ": %s()\n", __func__);
> > > > +}
> > > > +
> > >
> > > ehhh how on earth can this be right?
> >
> > oh and this stuff also violates the "one value per file" rule; can we
> > fix that urgently before it becomes part of the ABI in 2.6.16??
> 
> Ok, I'll admit to being a bit clueless about this.  I'm not familiar
> with the "one value per file" rule; can someone please explain?

sysfs consists of files that only have one value per file.  Please do
not do otherwise, as it will become a maintance nightmare over time.
See the documentation file that Randy pointed you at for details.

> On Sunday 05 March 2006 07:55, Greg KH wrote:
> > Ugh.  Good catch, it isn't right.  Gotta love it when people try to
> > ignore the helpful messages the kernel gives you when you use an API
> > wrong :(
> 
> Hmm... I don't recall seeing any messages from the kernel.  What
> are you seeing?

If you do not have a release function, you will see the messages.  Just
putting a printk() in a release function is not acceptable, you really
need to free the memory from it.  If not, then your code is really
wrong...

thanks,

greg k-h