netdev - Re: [PATCH] wifi: rt2x00: use explicitly signed type for clamping

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:   Wed, 19 Oct 2022 00:10:07 +0300
From:   Andy Shevchenko <andriy.shevchenko@...ux.intel.com>
To:     "Jason A. Donenfeld" <Jason@...c4.com>
Cc:     linux-wireless@...r.kernel.org, netdev@...r.kernel.org,
        linux-kernel@...r.kernel.org,
        Andrew Morton <akpm@...ux-foundation.org>,
        Stanislaw Gruszka <stf_xl@...pl>,
        Helmut Schaa <helmut.schaa@...glemail.com>,
        Kalle Valo <kvalo@...nel.org>
Subject: Re: [PATCH] wifi: rt2x00: use explicitly signed type for clamping

On Tue, Oct 18, 2022 at 02:58:30PM -0600, Jason A. Donenfeld wrote:
> On Tue, Oct 18, 2022 at 2:57 PM Andy Shevchenko
> <andriy.shevchenko@...ux.intel.com> wrote:
> > On Tue, Oct 18, 2022 at 02:52:43PM -0600, Jason A. Donenfeld wrote:
> > > On Tue, Oct 18, 2022 at 11:40:54PM +0300, Andy Shevchenko wrote:
> > > > On Tue, Oct 18, 2022 at 02:27:34PM -0600, Jason A. Donenfeld wrote:
> > > > > On some platforms, `char` is unsigned, which makes casting -7 to char
> > > > > overflow, which in turn makes the clamping operation bogus. Instead,
> > > > > deal with an explicit `s8` type, so that the comparison is always
> > > > > signed, and return an s8 result from the function as well. Note that
> > > > > this function's result is assigned to a `short`, which is always signed.
> > > >
> > > > Why not to use short? See my patch I just sent.
> > >
> > > Trying to have the most minimal change here that doesn't rock the boat.
> > > I'm not out to rewrite the driver. I don't know the original author's
> > > rationales. This patch here is correct and will generate the same code
> > > as before on architectures where it wasn't broken.
> > >
> > > However, if you want your "change the codegen" patch to be taken
> > > seriously, you should probably send it to the wireless maintainers like
> > > this one, and they can decide. Personally, I don't really care either
> > > way.
> >
> > I have checked the code paths there and I found no evidence that short can't be
> > used. That's why my patch.
> 
> Do you have a rationale why you want to change codegen?

It's not a hot path as far as I understand and keeping data types aligned seems
to me worth it even if codegen is changed. IS it so awful with short?

-- 
With Best Regards,
Andy Shevchenko