linux-hardening - Re: [PATCH 2/3] gcc-plugins/stackleak: Exactly match strings instead of prefixes

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <202202061046.040D6A0A0@keescook>
Date:   Sun, 6 Feb 2022 10:49:15 -0800
From:   Kees Cook <keescook@...omium.org>
To:     Linus Torvalds <torvalds@...ux-foundation.org>
Cc:     Alexander Popov <alex.popov@...ux.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        Josh Poimboeuf <jpoimboe@...hat.com>,
        Borislav Petkov <bp@...en8.de>,
        Masahiro Yamada <masahiroy@...nel.org>,
        linux-kernel@...r.kernel.org, linux-hardening@...r.kernel.org
Subject: Re: [PATCH 2/3] gcc-plugins/stackleak: Exactly match strings instead
 of prefixes

On Sun, Feb 06, 2022 at 10:34:11AM -0800, Linus Torvalds wrote:
> On Sun, Feb 6, 2022 at 9:45 AM Kees Cook <keescook@...omium.org> wrote:
> >
> > +       return !strncmp(TREE_STRING_POINTER(node), string, length);
> 
> Why is this "strncmp()"? That makes no sense when you've just checked
> the exact lengths of both sides.
> 
> You're not comparing strings any more, you've already checked the end
> of the string - you are comparing memory contents.
> 
> So make it just do a "memcmp()".

Yeah, good point. I'll change this for v2, pending more feedback.

> > +#define STRING_EQUAL(node, str)        string_equal(node, str, strlen(str))
> 
> .. and please change this name too, since it's not comparing two
> strings. The first argument is something else entirely.
> 
> It's checking the node value of a section, give it some name related to that.

Technically, yes. The naming bikeshed here is odd since it's called
"STRING" by gcc internals, and it *might* be a "C string", etc etc. I'll
rename it...

> I do also get the feeling that the nodes should actually be checked to
> be a STRING_CST rather than these blind TREE_VALUE() following things,
> but I don't really know the rules for gcc plugin internals very well -
> or at all, really.

I'll double-check this, but if it's not a STRING_CST something else has
gone very wrong already. But I'm a fan of robustness, so sure. :)

-Kees

-- 
Kees Cook