Subject: Re: Opinion on ordering of writel vs. stores to RAM
From: Benjamin Herrenschmidt <benh@kernel.crashing.org>
To: David Miller <davem@davemloft.net>
Cc: segher@kernel.crashing.org, jbarnes@virtuousgeek.org, jeff@garzik.org,
       paulus@samba.org, torvalds@osdl.org, linux-kernel@vger.kernel.org,
       akpm@osdl.org
In-Reply-To: <20060911.173208.74750403.davem@davemloft.net>
References: <D680AFCF-11EC-48AD-BBC2-B92521DF442A@kernel.crashing.org>
	 <20060911.062144.74719116.davem@davemloft.net>
	 <8DA3BCBF-0F19-4CF0-B22E-91E57E7CB033@kernel.crashing.org>
	 <20060911.173208.74750403.davem@davemloft.net>
Content-Type: text/plain
Date: Tue, 12 Sep 2006 10:49:07 +1000
Message-Id: <1158022147.15465.29.camel@localhost.localdomain>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 925
Lines: 23


> It's silly because if you just use different interface
> names for the different semantics, the caller can
> ask for what he wants at the call site and no conditionals
> are needed in the implementation.

As Paulus also pointed out, having writel() behave differently based on
some magic done earlier at map time makes it harder to understand what
happens when reading the code, and thus harder to audit drivers for
missing barriers etc... since it's not obvious at first sight wether a
driver is using ordered or relaxed semantics. Thus I prefer keeping two
speparate interfaces.

We've come up with the __writel() name, but I'm open to proposals for
something nicer :)

Ben.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/