Re: RFC: ACPI/scsi/libata integration and hotswap

From: Matthew Garrett <mjg59@srcf.ucam.org>
To: Jeff Garzik <jgarzik@pobox.com>
Cc: Christoph Hellwig <hch@infradead.org>,
	Alan Cox <alan@lxorguk.ukuu.org.uk>,
	randy_d_dunlap@linux.intel.com, linux-ide@vger.kernel.org,
	linux-scsi@vger.kernel.org, linux-kernel@vger.kernel.org,
	acpi-devel@lists.sourceforge.net
Subject: Re: RFC: ACPI/scsi/libata integration and hotswap
Date: Fri, 9 Dec 2005 12:11:25 +0000	[thread overview]
Message-ID: <20051209121124.GA25974@srcf.ucam.org> (raw)
In-Reply-To: <43997171.9060105@pobox.com>

On Fri, Dec 09, 2005 at 06:58:41AM -0500, Jeff Garzik wrote:

> If this is for hotswap, as I noted, libata doesn't need this at all.
> 
> If the hardware supports it, then libata will support it directly. 
> There is no ACPI-specific magic, because ACPI does nothing but talk to 
> the same hardware libata is talking to.

If libata knows how to talk to the random hardware attached to a Dell 
laptop hotswap bay, I'll be amazed. Ejecting the drive generates a 
system management interrupt, which then causes the ACPI code to check a 
register in a block of machine-specific registers and generate an ACPI 
notification. As far as I can tell, the controller has no say in the 
matter at all - the Intel specs seem to suggest that ICH6 doesn't 
generate a hotswap interrupt unless you're using AHCI (which this 
hardware doesn't).

So, as far as I can tell, there /is/ ACPI-specific magic on 
current-generation hardware. If we're lucky, they'll move to AHCI in 
future and implement things properly there - but I wouldn't count on it.

-- 
Matthew Garrett | mjg59@srcf.ucam.org