Re: CVE-2023-52734: net: sched: sch: Bounds check priority

[Michal Hocko <mhocko@suse.com>]

On Tue 28-05-24 21:06:39, Greg KH wrote:
> On Tue, May 28, 2024 at 09:53:12AM +0200, Michal Hocko wrote:
> > Is this really soemthing that should be getting a CVE assigned?
> > First the fix is incomplete - 9cec2aaffe96 ("net: sched: sch: Fix off by one in htb_activate_prios()")
> 
> Incomplete fixes are still part of a fix :)

Sigh

> > Second is this even real problem? https://lore.kernel.org/all/Y9V3mBmLUcrEdrTV@pop-os.localdomain/
> > suggests it is not.
> 
> Ah, good catch, I didn't see that.  I'll go revoke this as it's not
> doing anything.

Thanks!

I wish the CVE review process would catch something like that before
issuing a CVE for it.

> > And third, WARN_ONs are considered a real deal by CVE team because
> > somebody might be running with panic_on_warn. This patch adds one!
> 
> Yes, but if you can't hit that by anything from userspace, it's not an
> issue and just dead code.  We'll have to wait for a future syzbot report
> to prove that wrong :)

I am not judging the patch itself. It is maintainers who should decide
whether this is something they want to accept.

I am questioning the decision to make it a CVE. Because if that was a
real deal then WARN_ON is something kernel CNA is considering a CVE worth
problem! So a CVE has been filed with a fix that is CVE itself.
Seriously how could this pass through the CVE review process?
-- 
Michal Hocko
SUSE Labs