Re: [RFC PATCH next]module: Fix mod->mkobj.kobj potentially freed too early

[Li Zhong <zhong@linux.vnet.ibm.com>]

On Thu, 2013-08-22 at 16:30 +0930, Rusty Russell wrote:
> Greg KH <gregkh@linuxfoundation.org> writes:
> > On Wed, Aug 21, 2013 at 05:49:58PM +0800, Li Zhong wrote:
> > >  struct kobj_type module_ktype = {
> > > +	.release   =	module_kobj_release,
> > >  	.sysfs_ops =	&module_sysfs_ops,
> > >  };
> >
> > Wait, as there is no release function here for the kobject (a different
> > problem), why is the deferred release function causing any problems?
> > There is no release function to call, so what is causing the oops?
> 
> Because DEBUG_KOBJECT_RELEASE does the kobject_put() sometime later,
> which is what causes the oops.
> 
> Since kobjects don't have an owner field, AFAICT someone *could* grab
> one in a module which is unloading, then put it after unload.   So this
> fixes a real bug, albeit not one seen in the real world.
> 
> Applied,

Oh, thank you, Rusty. 

I just sent out another version... which fix it in another way as Greg
suggested, could you please also help to take a look at it?

Thanks, Zhong

> Rusty.
>