Re: [PATCH 09/18] dma-debug: add checking for map/unmap_page/single

[Joerg Roedel <joro@8bytes.org>]

On Thu, Mar 19, 2009 at 10:39:30AM +0900, FUJITA Tomonori wrote:
> On Fri, 6 Mar 2009 14:30:20 +0100
> Joerg Roedel <joerg.roedel@amd.com> wrote:
> > +	if (map_single) {
> > +		entry->type = dma_debug_single;
> > +		check_for_stack(dev, page_address(page) + offset);
> 
> Why you don't call check_for_stack() for dma_map_page()?
> 
> page_address(page) could be invalid with dma_map_page() so the check
> can be pointless. However, you call check_for_stack() with dma_map_sg,
> which the check can be pointless too with; I think that you call
> check_for_stack() in an inconsistent way.

I wasn't aware that sg mappings support highmem too. I did the check
only for map_single because I havn't found a way to check if the address
is mapped. Its not so important because the pointer is never
dereferenced but only compared. Looking again at it I think a check for
PageHighMem() should be sufficient to check if the page is mapped. I
will update the code. Thanks for pointing this out.

Joerg