phc-discussions - Re: [PHC] integer multiply was not always constant-time

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 for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <CAOLP8p5Cj2Xm3nY3CkO8my0DQiXGXhmbPJgmMbZdzfoDRjWG+g@mail.gmail.com>
Date: Tue, 4 Mar 2014 18:02:56 -0500
From: Bill Cox <waywardgeek@...il.com>
To: discussions@...sword-hashing.net
Subject: Re: [PHC] integer multiply was not always constant-time

On Tue, Mar 4, 2014 at 5:41 PM, Solar Designer <solar@...nwall.com> wrote:
> BTW, let's not forget that even MUL/IMUL were not constant time on 486
> and prior (and it's similar on similarly old non-x86 archs).  So if we
> care about being timing-safe even on CPUs this old, we should probably
> avoid multiplication in the portion where we avoid cache timing leaks.
>
> Alexander

I think multiply speed remains data-dependent on ARM Thumb
architectures (I've not checked which).  On pretty much all of these
older machines, I think setting the number of multiplies per loop to
zero is probably the right thing to do.  Not only are the multiplies
not constant time, but they take a bunch more cycles, and there's no
multiple-issue instruction unit, so this is all in-line with the
runtime, and will significantly reduce the memory we can fill.  It's
probably best just to fill memory as rapidly as possible on these
machines.

Bill