linux-kernel - Re: cleanup: Make no_free_ptr() __must

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]

Date:   Tue, 15 Aug 2023 15:53:39 +0200
From:   Peter Zijlstra <peterz@...radead.org>
To:     Rasmus Villemoes <linux@...musvillemoes.dk>
Cc:     Linus Torvalds <torvalds@...ux-foundation.org>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        dan.j.williams@...el.com, linux-kernel@...r.kernel.org,
        Nick Desaulniers <ndesaulniers@...gle.com>
Subject: Re: cleanup: Make no_free_ptr() __must_check

On Tue, Aug 15, 2023 at 01:28:37PM +0200, Rasmus Villemoes wrote:
> On 15/08/2023 12.52, Peter Zijlstra wrote:
> > 
> > recent discussion brought about the realization that it makes sense for
> > no_free_ptr() to have __must_check semantics in order to avoid leaking
> > the resource.
> > 
> 
> > +static inline __must_check void * __no_free_ptr(void **pp)
> > +{ void *p = *pp; *pp = NULL; return p; }
> > +
> >  #define no_free_ptr(p) \
> > -	({ __auto_type __ptr = (p); (p) = NULL; __ptr; })
> > +	(({ void * __maybe_unused ___t = (p); }), \
> > +	 ((typeof(p))__no_free_ptr((void **)&(p))))
> 
> So this does seem to work as advertised, but it could perhaps use some
> comments. Because at first I read this as one big statement expression,
> and I had this memory of a __must_check function call being the last
> expression in such had no effect at all [1]. But this is actually a
> comma expression.

Right, I can into that as well, that was infact the first thing I tried.
Most vexing indeed.

> 
> Also, isn't it more complicated than necessary? Can we get rid of the
> inner stmt expr and tmp var by just making it
> 
>   ((void) (p), ((typeof(p))__no_free_ptr((void **)&(p)))
> 
> which is more or less the whole reason comma expressions is a thing.

Ah, so the point of the statement expression before the comma is to
validate that (p) is in fact a pointer, and to that effect we assign it
to a 'void *' temporary.

If that case is invalid, we'll get a compile fail with a dodgy message.

I did this, because (void **)&(p) looses type integrity due to the cast.

But yeah, I suppose it all needs a wee comment.