Re: [RFC][PATCH] Prevent overriding of Symbols in the Kernel, avoiding Undefined behaviour

[Ashutosh Naik <ashutosh.naik@gmail.com>]

On 12/13/05, Alan Cox <alan@lxorguk.ukuu.org.uk> wrote:

> Its almost the 0% solution. The kernel as shipped doesn't seem to have
> any clashing symbols like this. The two sets of cases people report are
>
> 1. Out of tree modules
> 2. Reconfiguring, rebuilding something from kernel to module and not
> cleaning up
>
> A dep time solution might fix one of those but robustness here would be
> good, especially as once the installation is incorrect end users can
> often trigger hotplug loads that cause problems.

I agree with this.

I also would like to add that, the exported symbol may not always be
in the same module. Imagine if Module A is loaded and Module B would
export one symbol with the same symbol name as a symbol in Module A,
then the symbol exported by Module B would still go through. Now
Imagine if that symbol does something like a kmem_cache_create of an
existing cache!!

I feel this is a security loophole and preventing duplicate *exported*
symbols in the kernel, might just solve it.

Regards
Ashutosh