From mboxrd@z Thu Jan  1 00:00:00 1970
From: Andreas Herrmann <andreas.herrmann@caviumnetworks.com>
Subject: Re: [PATCH v2 09/13] MIPS: Add functions for hypervisor call
Date: Tue, 3 Jun 2014 17:03:37 +0200
Message-ID: <20140603150337.GA28045@alberich>
References: <1401313936-11867-1-git-send-email-andreas.herrmann@caviumnetworks.com>
 <1401313936-11867-10-git-send-email-andreas.herrmann@caviumnetworks.com>
 <20140603083031.GP17197@linux-mips.org>
Mime-Version: 1.0
Content-Type: text/plain; charset="utf-8"
Cc: <linux-mips@linux-mips.org>, David Daney <ddaney.cavm@gmail.com>,
	James Hogan <james.hogan@imgtec.com>, <kvm@vger.kernel.org>,
	David Daney <david.daney@cavium.com>
To: Ralf Baechle <ralf@linux-mips.org>
Return-path: <kvm-owner@vger.kernel.org>
Received: from mail-bl2lp0211.outbound.protection.outlook.com ([207.46.163.211]:59799
	"EHLO na01-bl2-obe.outbound.protection.outlook.com"
	rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP
	id S1753311AbaFCPEt (ORCPT <rfc822;kvm@vger.kernel.org>);
	Tue, 3 Jun 2014 11:04:49 -0400
Content-Disposition: inline
In-Reply-To: <20140603083031.GP17197@linux-mips.org>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

On Tue, Jun 03, 2014 at 10:30:31AM +0200, Ralf Baechle wrote:
> On Wed, May 28, 2014 at 11:52:12PM +0200, Andreas Herrmann wrote:
> 
> > +/*
> > + * Hypercalls for KVM.
> > + *
> > + * Hypercall number is passed in v0.
> > + * Return value will be placed in v0.
> > + * Up to 3 arguments are passed in a0, a1, and a2.
> > + */
> > +static inline unsigned long kvm_hypercall0(unsigned long num)
> > +{
> > +	register unsigned long n asm("v0");
> > +	register unsigned long r asm("v0");
> 
> Btw, is it safe to put two variables in the same register?

I think it's safe.

If we would have a matching constraint letter (say "v" for register v0) the
asm should translate to

        __asm__ __volatile__(
	       KVM_HYPERCALL
                : "=v" (n) : "v" (r) : "memory"
                );

which isn't unusual on other archs. (Or maybe I am just biased from
x86 ... or missed something else.)

> The syscall wrappers that used to be in <asm/unistd.h> were occasionally
> hitting problems which eventually forced me to stop forcing variables
> into particular registers instead using a MOVE instruction to shove
> each variable into the right place.
> 
> Of course they were being used from non-PIC and PIC code, kernel and userland
> so GCC had a much better chance to do evil than in the hypercall wrapper
> case - but it made me paranoid ...



Andreas