Re: linux-2.4.28 released

Re: linux-2.4.28 released

[Jeff Garzik]

On Thu, Nov 18, 2004 at 09:12:36AM -0200, Marcelo Tosatti wrote:
> On Thu, Nov 18, 2004 at 06:50:03PM +0900, Tomita, Haruo wrote:
> > Hi,
> > 
> > It seems that combined mode does not work at linux-2.4.28 about
> > the ata_piix driver of the Intel 82801EB/82801ER SATA controller
> > of Intel 82801EB/82801ER. In using combined mode, 
> > I think that the following patches are required. Is this right?
> 
> Yes, I think so? Jeff is the man.
> 
> I dislike the ____request_resource() hack, it has been rejected and 
> Jeff agreed with me here.

The reason ____request_resource() is used for combined mode is to
facilitate libata taking one PCI device, and IDE driver taking another
PCI device.  This is done because libata did not support PATA, and so,
could not drive the PATA controller.

Now PATA and ATAPI are working, so we could present the IMO ideal
situation:  libata can support combined mode best by supporting both
PATA and SATA controllers, on the ICH5.  That way DMA works for both
PATA and SATA (DMA doesn't work for PATA, in split-driver configuration),
and there is no split-driver to worry about.

I think there may need to be some code to prevent the IDE driver from
claiming the legacy ISA ports.

	Jeff

Re: linux-2.4.28 released

[Alan Cox]

On Fri, Nov 19, 2004 at 08:48:32AM -0500, Jeff Garzik wrote:
> PATA and SATA (DMA doesn't work for PATA, in split-driver configuration),
> and there is no split-driver to worry about.
> 
> I think there may need to be some code to prevent the IDE driver from
> claiming the legacy ISA ports.

Its called "request_resource". If you want the resource claim it. IDE will
be a good citizen.

Alan

Re: linux-2.4.28 released

[Jeff Garzik]

Alan Cox wrote:
> On Fri, Nov 19, 2004 at 08:48:32AM -0500, Jeff Garzik wrote:
> 
>>PATA and SATA (DMA doesn't work for PATA, in split-driver configuration),
>>and there is no split-driver to worry about.
>>
>>I think there may need to be some code to prevent the IDE driver from
>>claiming the legacy ISA ports.
> 
> 
> Its called "request_resource". If you want the resource claim it. IDE will
> be a good citizen.

That's what the quirk does.  libata still needs to find out who obtained 
the resource, not blindly grab it (and fail).

	Jeff

Re: linux-2.4.28 released

[Geert Uytterhoeven]

On Fri, 19 Nov 2004, Jeff Garzik wrote:
> Alan Cox wrote:
> > On Fri, Nov 19, 2004 at 08:48:32AM -0500, Jeff Garzik wrote:
> > > PATA and SATA (DMA doesn't work for PATA, in split-driver configuration),
> > > and there is no split-driver to worry about.
> > > 
> > > I think there may need to be some code to prevent the IDE driver from
> > > claiming the legacy ISA ports.
> > 
> > Its called "request_resource". If you want the resource claim it. IDE will
> > be a good citizen.
> 
> That's what the quirk does.  libata still needs to find out who obtained the
> resource, not blindly grab it (and fail).

If libata would be initialized before IDE, it could grab the resource during
probing.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds

RE: linux-2.4.28 released

[Tomita, Haruo]

Sorry for not writing you soon.

Jeff Garzik wrote:

> >>PATA and SATA (DMA doesn't work for PATA, in split-driver 
> configuration),
> >>and there is no split-driver to worry about.
> >>
> >>I think there may need to be some code to prevent the IDE 
> driver from
> >>claiming the legacy ISA ports.
> > 
> > 
> > Its called "request_resource". If you want the resource 
> claim it. IDE will
> > be a good citizen.
> 
> That's what the quirk does.  libata still needs to find out 
> who obtained the resource, not blindly grab it (and fail).

I also think so. 

It may be unavoidable one that ata_piix does not work. 
But, it is a problem that a DMA transfer does not enable by piix. 
Don't you think so?
--
Haruo

Re: linux-2.4.28 released

[Jeff Garzik]

Tomita, Haruo wrote:
> It may be unavoidable one that ata_piix does not work. 
> But, it is a problem that a DMA transfer does not enable by piix. 
> Don't you think so?


This is unavoidable.  Two drivers grabbing the same PCI I/O range is
dangerous.

Now that libata supports PATA, it would be easier to let libata support
both SATA and PATA.  Since that is a single driver, it makes DMA easy to
support for both SATA/PATA.

If libata does this, there needs to be a "ide=disable" or
"legacy_ide=libata" switch added somewhere, for the cases (most distros)
where IDE driver is built-in.

	Jeff