page_to_phys

page_to_phys

[Kristoffer Glembo]

Hi,

Currently in io_32.h page_to_phys is defined as:

#define page_to_phys(page)	(((page) - mem_map) << PAGE_SHIFT)

Should it not be:

#define page_to_phys(page)	((((page) - mem_map) << PAGE_SHIFT) + phys_base)

or perhaps:

#define page_to_phys(page)	((dma_addr_t)page_to_pfn(page) << PAGE_SHIFT)

Best regards,
Kristoffer Glembo

Re: page_to_phys

[David Miller]

From: Kristoffer Glembo <kristoffer@gaisler.com>
Date: Fri, 09 Oct 2009 16:02:14 +0200

> Currently in io_32.h page_to_phys is defined as:
> 
> #define page_to_phys(page)	(((page) - mem_map) << PAGE_SHIFT)
> 
> Should it not be:
> 
> #define page_to_phys(page) ((((page) - mem_map) << PAGE_SHIFT) +
> #phys_base)
> 
> or perhaps:
> 
> #define page_to_phys(page) ((dma_addr_t)page_to_pfn(page) <<
> #PAGE_SHIFT)

Sparc 32-bit doesn't offset the mem_map at phys_base.

What is offset by phys_base is what gets mapped at PAGE_OFFSET
virtually.

That's why the things that translate between virtual and physical
addresses do take phys_base into account.

Re: page_to_phys

[kristoffer]

Quoting David Miller <davem@davemloft.net>:

>
> Sparc 32-bit doesn't offset the mem_map at phys_base. 
>
> What is offset by phys_base is what gets mapped at PAGE_OFFSET
> virtually. 
>

Thanks for the clarification. Got temporarily confused. :-)

Best regards,
Kristoffer Glembo