Subject: Re: Memory barriers and spin_unlock safety
From: Alan Cox <alan@lxorguk.ukuu.org.uk>
To: Robert Hancock <hancockr@shaw.ca>
Cc: Linus Torvalds <torvalds@osdl.org>,
       linux-kernel <linux-kernel@vger.kernel.org>
In-Reply-To: <441225B7.5010003@shaw.ca>
References: <5Ml19-2Ki-19@gated-at.bofh.it> <5MlO0-3JU-13@gated-at.bofh.it>
	 <5MCF0-2TS-27@gated-at.bofh.it> <5MITJ-2l4-15@gated-at.bofh.it>
	 <5NXxl-6WZ-9@gated-at.bofh.it> <5NY0h-7wa-1@gated-at.bofh.it>
	 <441225B7.5010003@shaw.ca>
Content-Type: text/plain
Content-Transfer-Encoding: 7bit
Date: Sun, 12 Mar 2006 18:09:12 +0000
Message-Id: <1142186952.20056.8.camel@localhost.localdomain>
Mime-Version: 1.0
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 874
Lines: 17

On Gwe, 2006-03-10 at 19:19 -0600, Robert Hancock wrote:
> PCI I/O writes are allowed to be posted by the host bus bridge (not 
> other bridges) according to the PCI 2.2 spec. Maybe no x86 platform 
> actually does this, but it's allowed, so technically a device would need 
> to do a read in order to ensure that I/O writes have arrived at the 
> device as well.

Existing Linux drivers largely believe that PCI I/O cycles as opposed to
MMIO cycles are not posted. At least one MIPS platform that did post
them ended up ensuring PCI I/O cycle posting didn't occur to get a
running Linux system - so its quite a deep assumption.

-
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/