linux-kernel - Re: [RFC] saturate check_*

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: <20200805232208.GT6753@gate.crashing.org>
Date:   Wed, 5 Aug 2020 18:22:08 -0500
From:   Segher Boessenkool <segher@...nel.crashing.org>
To:     Rasmus Villemoes <linux@...musvillemoes.dk>
Cc:     Kees Cook <keescook@...omium.org>, Jason Gunthorpe <jgg@...pe.ca>,
        Leon Romanovsky <leon@...nel.org>,
        "Gustavo A. R. Silva" <gustavoars@...nel.org>,
        Matthew Wilcox <willy@...radead.org>,
        linux-kernel@...r.kernel.org, kernel-hardening@...ts.openwall.com
Subject: Re: [RFC] saturate check_*_overflow() output?

Hi Rasmus,

On Wed, Aug 05, 2020 at 01:38:58PM +0200, Rasmus Villemoes wrote:
> I'm guessing gcc has some internal very early simplification that
> replaces single-expression statement-exprs with just that expression,
> and the warn-unused-result triggers later. But as soon as the
> statement-expr becomes a little non-trivial (e.g. above), my guess is
> that the whole thing gets assigned to some internal "variable"
> representing the result, and that assignment then counts as a use of the
> return value from must_check_overflow() - cc'ing Segher, as he usually
> knows these details.

A statement expression is not a statement (it's an expression), which
turns half of the world upside down.  This GCC extension often has weird
(or at least non-intuitive) side effects, together with other extensions
(like attributes), etc.

This may be a convoluted way of saying "I don't know, look at c/c-decl.c
(and maybe c/c-parser.c) to see if you can find out" ;-)


> Anyway, we don't need to apply it to the last expression inside ({}), we
> can just pass the whole ({}) to must_check_overflow() as in

<snip>

Yes, much nicer :-)  Crisis averted, etc.


Segher