linux-kernel - Re: [PATCH 02/10] x86, asm: use bool for bitops and other assembly outputs

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: <d1f14d3c-b4dc-bfbc-bae9-90ffc9282139@zytor.com>
Date:	Wed, 8 Jun 2016 02:40:21 -0700
From:	"H. Peter Anvin" <hpa@...or.com>
To:	Ingo Molnar <mingo@...nel.org>
Cc:	Peter Zijlstra <peterz@...radead.org>,
	"H. Peter Anvin" <hpa@...ux.intel.com>,
	Thomas Gleixner <tglx@...utronix.de>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	Andy Lutomirski <luto@...capital.net>,
	Borislav Petkov <bp@...en8.de>,
	Linus Torvalds <torvalds@...ux-foundation.org>,
	Arnd Bergmann <arnd@...db.de>
Subject: Re: [PATCH 02/10] x86, asm: use bool for bitops and other assembly
 outputs

On 06/08/16 02:20, Ingo Molnar wrote:
> 
> Also, unless I'm missing something it's not really 'hard' or dangerous per se to 
> do that change for every architecture, just incredibly boring! ;-)
> 
> I'm not sure how much it matters though, given other asymmetries in the bitops API 
> signatures - does anyone have any preferences?
> 

My big concern is doing a change to another architectures which I can't
test... and I will have to make assumptions about the properties of
asm() in that architecture.

Furthermore, if/when other architectures get support for =@cc or an
equivalent, they probably want to change their stuff.

>> Perhaps I'm wrong, but I'd really like to avoid the upcasting to "int"
>> which isn't needed, because in my testing I find that it definitely
>> encourages gcc to generate poor code.
> 
> Yeah, absolutely. I hate 'bool' with a vengence but if 'int' generates worse code 
> with modern compilers then I'm not going to argue for worse code. Would a 'char' 
> return type be very weird?

Another reason to not hate on "bool" so much: I have personally gotten
bitten a few to many times by programmers who thought returning -1 or 2
for what was normally a 0/1 flag in some kind of exceptional case.  If
the variable is bool you know that that can't happen.

One thing, though: we should NOT use bool as input to a system call,
because coming from userspace we cannot enforce the required invariant
that bits[7:1] == 0.