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

[Russell King <rmk+lkml@arm.linux.org.uk>]

On Sun, Mar 21, 2004 at 03:11:58PM -0800, Linus Torvalds wrote:
> On Sun, 21 Mar 2004, Russell King wrote:
> > Remember that we're fond of telling driver writers to use scatter gather
> > lists rather than grabbing one large contiguous memory chunk...  So
> > they did exactly as we told them.  Using pci_alloc_consistent and/or
> > dma_alloc_coherent and built their own scatter lists.
> 
> I do think that we should introduce a "map_dma_coherent()" thing, which 
> basically takes a list of pages that have been allocated by 
> dma_alloc_coherent(), and remaps them into user space. How hard can that 
> be?
> 
> In fact, on a lot of architectures (well, at least x86, and likely
> anything else that doesn't use any IOTLB and just allocates a chunk of
> physical memory), I think the "map_dma_coherent()" thing should basically
> just become a "remap_page_range()". Ie something like
> 
> 	#define map_dma_coherent(vma, vaddr, len) \
> 		remap_page_range(vma, vma->vm_start, __pa(vaddr), len, vma->vm_page_prot)
> 
> for the simple case.

Ok, splitting hairs, for the coherent contiguous case, what about:

int dma_coherent_map(struct vm_area_struct *vma, void *cpu_addr,
		     dma_addr_t dma_addr, size_t size);

and x86 would be:

#define dma_coherent_map(vma,cpu_addr,dma_addr,size)	\
	remap_page_range(vma, vma->vm_start, __pa(cpu_addr), \
			 size, vma->vm_page_prot)

This then leaves the PCI BAR case and the DMA coherent SG buffer case,
though neither of those fall within my personal problem space at present.

-- 
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 PCMCIA      - http://pcmcia.arm.linux.org.uk/
                 2.6 Serial core