RE: [PATCH 2/7] iommu/amd: Implement DOMAIN_ATTR_GEOMETRY attribute

From: Sethi Varun-B16395
Date: Wed Jul 11 2012 - 09:28:42 EST




> -----Original Message-----
> From: Joerg Roedel [mailto:joerg.roedel@xxxxxxx]
> Sent: Wednesday, July 11, 2012 5:47 PM
> To: Sethi Varun-B16395
> Cc: iommu@xxxxxxxxxxxxxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH 2/7] iommu/amd: Implement DOMAIN_ATTR_GEOMETRY
> attribute
>
> Hi Sethi,
>
> On Wed, Jul 11, 2012 at 11:47:39AM +0000, Sethi Varun-B16395 wrote:
> > I am currently working on upstreaming the Freescale IOMMU driver. As
> > discussed (with Scott Wood) a while back, this representation of the
> > GEOMETRY attribute doesn't go well with our IOMMU implementation. Our
> > representation of the geometry would be more complex and is certainly
> > modifiable. Our IOMMU allows for creating custom DMA mappings by
> > defining a set of DMA sub windows within a given window. We need a
> > mechanism for specifying the number of DMA subwindows as a part of the
> > domain geometry. My suggestion would be to make the GEOMETRY attribute
> > implementation dependent
>
> Yeah, we had some discussions in the past and I came to the conclusion
> that its best to wait for your code and decide then if and how to change
> the interface to better suit your needs. The interface is not exposed to
> userspace so we can easily change it.
>
> People are waiting for this feature and I don't want to delay it further
> unless it is absolutly necessary.
The API call for setting domain attributes is an essential part of our implementation and
GEOMETRY is a critical attribute for us. I am currently working on our driver and hopefully
should be able to submit it in the next two weeks. In my case I was planning on letting the
geometry attribute be implementation dependent.

-Varun

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