Re: [PATCH] kobject: Access kobject name with caution if state is not initialized

[Dmitry Torokhov]

On Mon, Aug 20, 2018 at 12:22 PM Greg Kroah-Hartman
<gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>
> On Mon, Aug 20, 2018 at 10:39:47PM +0530, Srikar Dronamraju wrote:
> > If kobject state is not initialized, then its not even certain that
> > kobject'name is initialized. Hence when accessing the kobject's name
> > tread carefully.
> >
> > A stupid module test like
> > https://github.com/srikard/tests/blob/master/modules/kobject_test.c
> > can panic the system.
>
> Lots of stupid modules can do dumb things.  Just don't do that.  The
> kernel is not built to keep you from doing stupid things in kernel code
> :)
>
> So I fail to see why this patch is needed.  What in-kernel code path is
> trying to print a kobject's name before it is initialized?  Why not fix
> that obvious bug instead of forcing the kernel core to protect from
> stupid code?

Kay decided to add some guards in:

commit 0f4dafc0563c6c49e17fe14b3f5f356e4c4b8806
Author: Kay Sievers <kay.sievers@xxxxxxxx>
Date:   Wed Dec 19 01:40:42 2007 +0100

   Kobject: auto-cleanup on final unref
...

+       if (!kobj->state_initialized) {
+               printk(KERN_ERR "kobject '%s' (%p): tried to add an "
+                      "uninitialized object, something is seriously wrong.\n",
+                      kobject_name(kobj), kobj);
+               dump_stack();
+               return -EINVAL;

Given that we have dump_stack() we can probably simply drop
kobject_name(kobj) instead of building even more elaborate checks. Or
just drop the whole check. Adding kobjects is somewhat uncommon
operation, plus "gabage in, garbage out".

Thanks.

-- 
Dmitry