Received-SPF: pass (google.com: domain of linux-crypto-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Thu, 5 May 2022 05:38:58 +0900
From:   Stafford Horne <shorne@gmail.com>
To:     Linus Torvalds <torvalds@linux-foundation.org>
Cc:     "Jason A. Donenfeld" <Jason@zx2c4.com>,
        Andy Shevchenko <andriy.shevchenko@intel.com>,
        Mikulas Patocka <mpatocka@redhat.com>,
        Andy Shevchenko <andy@kernel.org>,
        device-mapper development <dm-devel@redhat.com>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Linux Crypto Mailing List <linux-crypto@vger.kernel.org>,
        Herbert Xu <herbert@gondor.apana.org.au>,
        "David S. Miller" <davem@davemloft.net>,
        Mike Snitzer <msnitzer@redhat.com>,
        Mimi Zohar <zohar@linux.ibm.com>,
        Milan Broz <gmazyland@gmail.com>
Subject: Re: [PATCH v2] hex2bin: make the function hex_to_bin constant-time
Message-ID: <YnLkYjOF2vEOdjOo@antec>
References: <alpine.LRH.2.02.2204250723120.26714@file01.intranet.prod.int.rdu2.redhat.com>
 <YnI7hE4cIfjsdKSF@antec>
 <YnJI4Ru0AlUgrr9C@zx2c4.com>
 <YnJOCbLtdATzC+jn@zx2c4.com>
 <YnJQXr3igEMTqY3+@smile.fi.intel.com>
 <YnJSQ3jJyvhmIstD@zx2c4.com>
 <CAHk-=wgb_eBdjM_mzEvXfRG2EhrSK5MHNGyAj7=4vxvN4U9Rug@mail.gmail.com>
 <CAHmME9q_-nfGxp8_VCqaritm4N8v8g67AzRjXs9du846JhhpoQ@mail.gmail.com>
 <CAAfxs77yaLvWx9KnkDZX7E1eDm9N-NVJn5n8=mCK9BU-cSob=A@mail.gmail.com>
 <CAHk-=wjLRo-6PbhbvMUDojbMo=L+2jc5VpCYTyF-LGxZPhUngA@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CAHk-=wjLRo-6PbhbvMUDojbMo=L+2jc5VpCYTyF-LGxZPhUngA@mail.gmail.com>
Precedence: bulk

On Wed, May 04, 2022 at 01:10:03PM -0700, Linus Torvalds wrote:
> On Wed, May 4, 2022 at 12:58 PM Stafford Horne <shorne@gmail.com> wrote:
> >
> > I have uploaded a diff I created here:
> >   https://gist.github.com/54334556f2907104cd12374872a0597c
> >
> > It shows the same output.
> 
> In hex_to_bin itself it seems to only be a difference due to some
> register allocation (r19 and r3 switched around).
> 
> But then it gets inlined into hex2bin and there changes there seem to
> be about instruction and basic block scheduling, so it's a lot harder
> to see what's going on.
> 
> And a lot of constant changes, which honestly look just like code code
> moved around by 16 bytes and offsets changed due to that.
> 
> So I doubt it's hex_to_bin() that is causing problems, I think it's
> purely code movement. Which explains why adding a nop or a fake printk
> fixes things.
> 
> Some alignment assumption that got broken?

This is what it looks like to me too.  I will have to do a deep dive on what is
going on with this particular build combination as I can't figure out what it is
off the top of my head.

This test is using a gcc 11 compiler, I tried with my gcc 12 toolchain and the
issue cannot be reproduced.

  - musl gcc 11 - https://musl.cc/or1k-linux-musl-cross.tgz
  - openrisc gcc 12 - https://github.com/openrisc/or1k-gcc/releases/tag/or1k-12.0.1-20220210-20220304

But again the difference between the two compiler outputs is a lot of register
allocation and offsets changes.  Its not easy to see anything that stands out.
I checked the change log for the openrisc specific changes from gcc 11 to gcc
12.  Nothing seems to stand out, mcount profiler fix for PIC, a new large binary
link flag.

-Stafford