Re: [PATCH v3 3/3] list_debug: Introduce CONFIG_DEBUG_LIST_MINIMAL

From: Kees Cook
Date: Thu Aug 10 2023 - 16:12:07 EST


On Wed, Aug 09, 2023 at 06:32:37PM +0200, Marco Elver wrote:
> On Wed, Aug 09, 2023 at 11:30AM -0400, Steven Rostedt wrote:
> [...]
> >
> > I would actually prefer DEBUG_LIST to select HARDEN_LIST and not the other
> > way around. It logically doesn't make sense that HARDEN_LIST would select
> > DEBUG_LIST. That is, I could by default want HARDEN_LIST always on, but not
> > DEBUG_LIST (because who knows, it may add other features I don't want). But
> > then, I may have stumbled over something and want more info, and enable
> > DEBUG_LIST (while still having HARDEN_LIST) enabled.
> >
> > I think you are looking at this from an implementation perspective and not
> > the normal developer one.
> >
> [...]
> >
> > That is, if DEBUG_LIST is enabled, we always call the
> > __list_add_valid_or_report(), but if only HARDEN_LIST is enabled, then we
> > do the shortcut.
>
> Good point - I think this is better. See below tentative v4.
>
> Kees: Does that also look more like what you had in mind?

Yeah, this looks good. My only nit would be a naming one. All the
other hardening features are named "HARDENED", but perhaps the "ED"
is redundant in the others. Still, consistency seems nicer. What do you
think of CONFIG_LIST_HARDENED ? (The modern trend for Kconfig naming tends
to keep the subsystem name first and then apply optional elements after.)

One note: do the LKDTM list hardening tests still pass? i.e.
CORRUPT_LIST_ADD
CORRUPT_LIST_DEL

> [...]
> + /*
> + * With the hardening version, elide checking if next and prev
> + * are NULL, LIST_POISON1 or LIST_POISON2, since the immediate
> + * dereference of them below would result in a fault.
> + */
> + if (likely(prev->next == entry && next->prev == entry))
> + return true;

I'm not super excited about skipping those checks, since they are
values that can be reached through kernel list management confusion. If
an attacker is using a system where the zero-page has been mapped
and is accessible (i.e. lacking SMAP etc), then attacks could still
be constructed. However, I do recognize this chain of exploitation
prerequisites is getting rather long, so probably this is a reasonable
trade off on modern systems.

-Kees

--
Kees Cook