Re: [PATCH 2.6.15-rc4 1/1] cpia: use vm_insert_page() instead of remap_pfn_range()

[Nick Holloway <Nick.Holloway@pyrites.org.uk>]

On Tue, Dec 06, 2005 at 07:10:12PM +0000, Christoph Hellwig wrote:
> >     pos = (unsigned long)(cam->frame_buf);
> >     while (size > 0) {
> > -           page = vmalloc_to_pfn((void *)pos);
> > -           if (remap_pfn_range(vma, start, page, PAGE_SIZE, PAGE_SHARED)) {
> > +           page = vmalloc_to_page((void *)pos);
> > +           if (vm_insert_page(vma, start, page)) {
>
> it would be nicer to do the arithmetis on pos as pointers rather than unsigned
> long.  Also you might want to use alloc_pages + vmap instead of vmalloc so that
> you already have a page array.  Or we should provide a helper that walks over
> a vmalloc'ed region and calls vmalloc_to_page + vm_insert_page.  Either way
> this type of code is duplicated far too much and we'd really need some better
> interface for it.

As I said in my previous mail, the patch was just switching to use
vm_insert_page, and not any other cleanups.

I agree that a helper is a good idea, as the vmalloc, SetPageReserved,
remap_pfn_range (was remap_page_range in 2.4) pattern has been copied
and pasted across many video4linux drivers.

The cpia driver could do with other cleanups.

        - It doesn't have a sysfs release callback (so says warning printk).
	- The colourspace conversion has been disabled, but should be
	  ripped out.
	- Needs to support V4L2 API

-- 
 `O O'  | Nick.Holloway@pyrites.org.uk
// ^ \\ | http://www.pyrites.org.uk/