Re: [PATCH] zfcp: kfree patch

[Greg KH <greg@kroah.com>]

On Mon, Jan 10, 2005 at 09:10:21AM +0100, Heiko Carstens wrote:
> > > > As nothing has changed in the patch, I still object to it. Actually,
> > > > krefs will BUG() out if you try to do this, so I might as well add a
> > > > patch so that kobjects do the same thing...
> > > So what's the fix for this generic (non-zfcp specific) problem then?
> > As no other code uses this type of "solution" I don't see how you can
> > say it is "generic".
> 
> Since you still don't like the patch I would like to know how it should
> look like that you would accept it.

As I don't know the scsi subsystem, I can only rely on the fact that
kfree should never be used as a release call.  So fix that, and I'm
happy :)

> Also in my original post I described what the problem is. If that problem
> exists then it is generic.
> Please see also "Module unloading in a reference counted world"
> -> http://lwn.net/Articles/67421/

I understand the "generic" issue, however no other subsystem needs to
use kfree as a release call.  Now either everyone else is broken or...

thanks,

greg k-h