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

[William Lee Irwin III <wli@holomorphy.com>]

On Sat, Mar 20, 2004 at 08:44:44PM +0100, Jaroslav Kysela wrote:
>> Yes, I'm sorry about that, but the ->nopage usage was requested by Jeff
>> Garzik and we're not gurus for the VM stuff. Because we're probably first
>> starting using of this mapping scheme, it resulted to problems.

On Sat, Mar 20, 2004 at 10:23:41PM +0000, Russell King wrote:
> Well, I've been told to effectively screw my idea by David Woodhouse,
> so may I make the radical suggestion that rm -rf linux/sound would
> also fix the problem.  No, didn't think that was acceptable either.
> Ok, so, how the fsck do we fix the sound drivers?  How do we mmap()
> memory provided by dma_alloc_coherent() into user space portably?
> It appears from what David Woodhouse has been going on about, even
> providing an architecture dma_coherent_to_page() interface isn't
> acceptable.
> If we can't answer that question, we might as well remove ALSA and
> OSS from the kernel because they are abusing existing kernel
> interfaces in ways which can not be solved.

Is there any possibility of an extension to remap_area_pages() that
could resolve this? I can't say I fully understood and/or remember
the issue with it that you pointed out.


-- wli