Re: Part 2: System Structure

[Jonathan S. Shapiro]

On Thu, 2006-05-25 at 15:53 +0200, Michal Suchanek wrote:
> I was somewhat concerned by the driver case. But if the dma buffers
> aren't too large they can be paid by the driver, and it would just map
> the user data in the buffer and return an empty page.

This is a lot more complicated to do securely than it looks, because the
exchanged page must first be zeroed (at which point a copy is just as
fast).

We looked at this kind of issue in our work on defensible networks
stacks. Dealing with DMA across a non-symmetric trust relationship is
very subtle, and very hard to get right.

And of course, this is the normal case in any high-efficiency driver.

shap