linux-kernel - Re: [PATCH] [RFC] x86: avoid -mtune=atom for objtool warnings

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]

Message-ID: <20170302144639.GA8969@gmail.com>
Date:   Thu, 2 Mar 2017 15:46:39 +0100
From:   Ingo Molnar <mingo@...nel.org>
To:     Josh Poimboeuf <jpoimboe@...hat.com>
Cc:     Arnd Bergmann <arnd@...db.de>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>,
        "H. Peter Anvin" <hpa@...or.com>, x86@...nel.org,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Denys Vlasenko <dvlasenk@...hat.com>
Subject: Re: [PATCH] [RFC] x86: avoid -mtune=atom for objtool warnings


* Josh Poimboeuf <jpoimboe@...hat.com> wrote:

> > > > Well, technically an invalid opcode is shorter code than generating an 
> > > > (integer) division by zero exception, right?
> > > 
> > > What does that matter if it's the wrong behavior?
> > 
> > Well, both terminate the program, and it's obvious if you look at it with a 
> > debugger what happened, right?
> 
> If it were obvious, we wouldn't be having this discussion :-)

Touche ;-)

> The only thing obvious to me was that gcc mysteriously removed a bunch of code 
> and replaced it with a 'ud2' instruction in the middle of the function for no 
> apparent reason.

I don't know what their motivation was, but if it's not a bug, if it was done 
intentionally, then I'd guess it's roughly the argument I made: in simple 
testcases it can be argued to be a code size improvement, plus it's probably 
allowed by the letter of the compiler standards (program termination behavior is 
notoriously platform dependent and thus vaguely specified) - but for real-life 
code I very much agree that it's a step backward in generated code quality...

Thanks,

	Ingo