non-PCI based libata-SATA driver

non-PCI based libata-SATA driver

[Deven Balani]

Hi all

I'm working on writing a non-PCI based libata-SATA driver in
linux-2.6.x for ARM based chipsets.

Can anybody suggest me a reference-code for a non-PCI based
libata-SATA driver for  2.6.x kernels ?

Any suggestions/comments will be of great help.

Thanks in advance.

Deven

Re: non-PCI based libata-SATA driver

[Alan Cox]

On Mer, 2006-02-22 at 17:34 +0530, Deven Balani wrote:
> Hi all
> 
> I'm working on writing a non-PCI based libata-SATA driver in
> linux-2.6.x for ARM based chipsets.
> 
> Can anybody suggest me a reference-code for a non-PCI based
> libata-SATA driver for  2.6.x kernels ?

If you look at http://zeniv.linux.org.uk/~alan/IDE you'll find the PATA
patches I did include some VLB and ISA devices. Providing you've got a
struct device * for whatever your SATA controller is (ie a bus model of
some sort) it is sane. If you don't then you need a couple of small
patches to allow for this, or to add device model stuff to them first. I
can send you those if need be and they are in the patches but not
seperated out.



Alan

Re: non-PCI based libata-SATA driver

[Deven Balani]

> If you look at http://zeniv.linux.org.uk/~alan/IDE you'll find the PATA
> patches I did include some VLB and ISA devices.
Thanks a lot. I'm going through these patches and will get back to you
in case of any problem.

I have a SATA Controller that is attached to HSX bus on ARM based
chipset. On this side, I'm exploring the feasibilty of coming up with
low-level libata-SATA driver (as libata is being said to support
generic bus interface). I've gone through the SATA drivers in kernel
code but all seem to be PCI specific.

Also, Is there any slight possiblity of doing the same with 2.4.x
kernels (by patching the kernel or whatever)?

Any suggestion/comments/advice will be of great help.

Thanks in advance.

Deven

On 2/23/06, Alan Cox <alan@lxorguk.ukuu.org.uk> wrote:
> On Mer, 2006-02-22 at 17:34 +0530, Deven Balani wrote:
> > Hi all
> >
> > I'm working on writing a non-PCI based libata-SATA driver in
> > linux-2.6.x for ARM based chipsets.
> >
> > Can anybody suggest me a reference-code for a non-PCI based
> > libata-SATA driver for  2.6.x kernels ?
>
> If you look at http://zeniv.linux.org.uk/~alan/IDE you'll find the PATA
> patches I did include some VLB and ISA devices. Providing you've got a
> struct device * for whatever your SATA controller is (ie a bus model of
> some sort) it is sane. If you don't then you need a couple of small
> patches to allow for this, or to add device model stuff to them first. I
> can send you those if need be and they are in the patches but not
> seperated out.
>
>
>
> Alan
>
>


--
"A smile confuses an approaching frown..."

Re: non-PCI based libata-SATA driver

[Jeff Garzik]

Deven Balani wrote:
>>If you look at http://zeniv.linux.org.uk/~alan/IDE you'll find the PATA
>>patches I did include some VLB and ISA devices.
> 
> Thanks a lot. I'm going through these patches and will get back to you
> in case of any problem.
> 
> I have a SATA Controller that is attached to HSX bus on ARM based
> chipset. On this side, I'm exploring the feasibilty of coming up with
> low-level libata-SATA driver (as libata is being said to support
> generic bus interface). I've gone through the SATA drivers in kernel
> code but all seem to be PCI specific.
> 
> Also, Is there any slight possiblity of doing the same with 2.4.x
> kernels (by patching the kernel or whatever)?

non-PCI works quite easily under 2.6.x, because libata core uses the 
generic DMA mapping lib.  It's already confirmed to work on at least one 
other ARM embedded chipset (sorry can't give more detail).

2.4.x is a lot of work to do non-PCI, largely because you have to deal 
with the lack of a generic DMA interface.  2.4.x libata is hardcoded to 
use PCI DMA mapping.

	Jeff

Re: non-PCI based libata-SATA driver

[Deven Balani]

> non-PCI works quite easily under 2.6.x, because libata core uses the
> generic DMA mapping lib.  It's already confirmed to work on at least one
> other ARM embedded chipset (sorry can't give more detail).
Thanks a lot for the valuable info.

> 2.4.x is a lot of work to do non-PCI, largely because you have to deal
> with the lack of a generic DMA interface.  2.4.x libata is hardcoded to
> use PCI DMA mapping.
At the back of my mind, i'm still exploring on how much work will be
required to do non-PCI libata-SATA in 2.4.x.
{I mean the changes that need to be done in kernel-2.4.23 onwards to
support non-PCI libATA-SATA. By some means, I have access to
physically contiguous memory blocks which can be used to do
DMA-ScatterGather.}

A pointer from your side will help me in hitting the nail right,
otherwise I'll be hurting myself a lot later on :)

Thanks a lot once again.

Deven