[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <aI_iNQXxU6wKPEN8@gate>
Date: Sun, 3 Aug 2025 17:27:01 -0500
From: Segher Boessenkool <segher@...nel.crashing.org>
To: Eric Biggers <ebiggers@...nel.org>
Cc: linux-crypto@...r.kernel.org, linux-kernel@...r.kernel.org,
Ard Biesheuvel <ardb@...nel.org>,
"Jason A . Donenfeld" <Jason@...c4.com>, linux-mips@...r.kernel.org,
linuxppc-dev@...ts.ozlabs.org, sparclinux@...r.kernel.org
Subject: Re: [PATCH 3/7] crypto: powerpc/md5 - Remove PowerPC optimized MD5
code
On Sun, Aug 03, 2025 at 03:14:38PM -0700, Eric Biggers wrote:
> On Sun, Aug 03, 2025 at 05:07:10PM -0500, Segher Boessenkool wrote:
> > On Sun, Aug 03, 2025 at 01:44:29PM -0700, Eric Biggers wrote:
> > > MD5 is insecure,
> >
> > Really? Have you found an attack? Can you explain it to the rest of
> > the world?
> >
> > MD5 is not recommended for future cryptographic purposes, there have
> > been some collission "attacks" on it (quotes because such a thing is
> > never an attack at all, merely an indication that not all is well with
> > it, somewhere in the future an actual vulnerability might be found).
> >
> > Since there are newer, better, *cheaper* alternatives available, of
> > course you should not use MD5 for anything new anymore. But claiming it
> > is insecure is FUD.
>
> Many attacks, including practical attacks, have been found on MD5 over
> the past few decades. Check out https://en.wikipedia.org/wiki/MD5
There is no new information on that page. There are no practical
attacks mentioned there, either, just some collission things (which
never can be practical attacks for most applications).
> > > This commit removes the PowerPC optimized MD5 code.
> >
> > Why? It would help to have real arguments for it!
>
> Sure, check out the commit message which mentioned multiple reasons why
> maintaining this code is not worthwhile.
Of course I have read that, but that information went missing, if you
intended to provide it :-(
You are replacing a known-working target implementation by a lower
performance generic implementation. But is that one known-working at
all? Does it come with tests? Was it tested to have the same outputs
as the existing thing, maybe? Just on a few inputs maybe.
We were not told anything like that.
Segher
Powered by blists - more mailing lists