Re: can device drivers return non-ram via vm_ops->nopage?

[David Woodhouse <dwmw2@infradead.org>]

On Sun, 2004-03-21 at 13:53 -0800, Linus Torvalds wrote:
> So I really put my veto on "nopage()" returning a PFN. That's just wrong, 
> wrong, wrong. It returns a "struct page" pointer, and it has lots of 
> reasons for that.

That's fine -- I wasn't suggesting nopage() should return a PFN.

I was suggesting that if someone wants to map something they're given by
dma_alloc_coherent() into memory, they should be given a PFN to deal
with -- _not_ a "struct page". Therefore, you can't use nopage() for
mapping dma_coherent memory into userspace.

Basically, we should consider the stuff returned by dma_alloc_coherent
to be 'non-RAM' in the context of your previous statement:
	'If a driver wants to map non-RAM pages, that's perfectly ok,
	but it MUST NOT happen through "nopage()".'

There are machines where you _cannot_ sensibly use host memory for
dma_coherent() allocations, but on which there _is_ a few megabytes of
SRAM hanging off the PCI bus which was put there specifically for that
purpose. So dma_alloc_coherent() returns something for which there is
not a valid 'struct page'.

-- 
dwmw2