linux-kernel - Re: [PATCH] coccinelle: api: add kzfree script

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

Message-ID: <alpine.DEB.2.21.2006041749520.2577@hadrien>
Date:   Thu, 4 Jun 2020 17:51:19 +0200 (CEST)
From:   Julia Lawall <julia.lawall@...ia.fr>
To:     Denis Efremov <efremov@...ux.com>
cc:     Julia Lawall <julia.lawall@...ia.fr>, cocci@...teme.lip6.fr,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] coccinelle: api: add kzfree script



On Thu, 4 Jun 2020, Denis Efremov wrote:

>
>
> On 6/4/20 5:15 PM, Julia Lawall wrote:
> > Did you try ... here but find that some subexpressions of E could be
> > modified in between?
>
> Yes, I tried to use "... when != E = E1 when != &E" and results were bad.
> Now, I've tried forall and when strict. Here are examples:
>
> // forall added
> // Works well, suitable for v2. One additional catch in w1 driver.
> @r depends on !patch && !(file in "lib/test_kasan.c") && !(file in "mm/slab_common.c") forall@
> expression *E; // pointer. Results are equal as if we use E.
> position p;
> @@
>
> * memset(E, 0, ...);
>   ... when != E // Is it enough to match &E, E = E1?

Yes.

> * kfree(E)@p;
>
> //no forall, when strict
> //results are bad, too many false positives
> @r depends on !patch && !(file in "lib/test_kasan.c") && !(file in "mm/slab_common.c")@
> expression *E;
> position p;
> @@
>
> * memset(E, 0, ...);
>   ... when != E // E is not enough here
>       when strict

OK, it's reassonable.

> * kfree(E)@p;
>
> I guess that the difference is that "forall" requires that whole pattern should occur on
> every path, "when strict" states that kfree should be called on every path after memset.
> This results in missed uses of E in loops and under conditions. How can I state in this
> case that E should not occur at all (in all paths) in between memset, kfree even as a
> subexpression?
>
> // Doesn't work well
>   ... when != E
>       when != if (...) { ... E ... }
>       when != for(...;...;...) { ... E ... }

Could you send an example of some C code on which the result is not
suitable?

julia