From mboxrd@z Thu Jan  1 00:00:00 1970
From: will.deacon@arm.com (Will Deacon)
Date: Thu, 5 Sep 2013 10:08:42 +0100
Subject: [PATCH v3 1/3] ARM: Introduce atomic MMIO modify
In-Reply-To: <522847FC.1080005@free-electrons.com>
References: <1377253445-2842-2-git-send-email-ezequiel.garcia@free-electrons.com>
 <20130823103802.GO4600@tarshish> <20130823110748.GD2389@localhost>
 <20130823113225.GF2389@localhost> <20130823114805.GF10971@arm.com>
 <20130830090807.GB25628@mudshark.cambridge.arm.com>
 <20130830091534.GA62188@MacBook-Pro.local>
 <20130830092032.GD25628@mudshark.cambridge.arm.com>
 <20130830100342.GD62188@MacBook-Pro.local>
 <522847FC.1080005@free-electrons.com>
Message-ID: <20130905090841.GA9696@mudshark.cambridge.arm.com>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Thu, Sep 05, 2013 at 09:59:40AM +0100, Gregory CLEMENT wrote:
> Hi all,

Hi Gregory,

> On 30/08/2013 12:03, Catalin Marinas wrote:
> > On Fri, Aug 30, 2013 at 10:20:33AM +0100, Will Deacon wrote:
> >> Perhaps we should just bit the bullet and define relaxed accessors for all
> >> architectures? It's not difficult to default them to the non-relaxed
> >> variants if the architecture doesn't provide an optimised implementation.
> > 
> > Yes, an asm-generic default relaxed would be good (that's what I
> > suggested earlier in this thread and it was discussed in the past). But
> > no-one volunteered ;).
> > 
> 
> I would like to make the things move on about this subject. Should it
> be possible to merge this version of the patch set? Currently the
> only users of this new API are drivers for ARM SoCs.

In the short term then, I'd keep this restricted to ARM. Without relaxed
accessors available across all architectures, I don't think we can make this
usefully generic (a readX/writeX version would be horribly slow on ARM).

> In the meantime, I am willing to introduce an asm-generic default
> relaxed variant of read and write, but as Catalin had already pointed
> in the past
> http://thread.gmane.org/gmane.linux.ports.arm.kernel/117626
> it may take time to get an agreement from all the other architectures.

I actually wrote those patches yesterday but, in doing so, I realised that
the kernel doesn't have well-defined semantics for relaxed I/O accessors
(i.e. different architectures use them to mean different things). That makes
portability is an issue, even if the functions are defined everywhere.

> That's why I propose that this patch set do not depend on the
> introduction of a asm-generic default relaxed variant of read and
> write. Later when it will be accepted then this new API will be moved
> in the asm-generic part.

Indeed, we can't make this generic until we've resolved the issues I mention
above. I'll start a new thread around that once I've put my thoughts
together, but that needn't block this patch imo.

That said, I'd still suggest resending a final version of this patch, since
there were some minor issues with this iteration.

Cheers,

Will