On 7/22/05, mark brown <markbrowni1000@xxxxxxxxx> wrote:
>
> Mulyadi Santosa <mulyadi.santosa@xxxxxxxxx> wrote:
>
> AFAIK, user level mem. requests is deferred until they
> are actually needed, at least Understanding the Linux
> Kernel book says so...
>
Yes thats true for user level memory requests ...
> > yes, IMHO this is correct. Like Fawaad said, memory
> > area that is mapped into kernel address space (above PAGE_OFFSET) always
> > has same virtual address no matter what context you are in (process
> > context, interrupt context. In this case, since module is also a part
> > of kernel, any allocation done via kmalloc is also mapped on kernel
> > address space, thus has constant virtual address on every situation.
>
> Can you give me a pointer in the kernel sources? I
> think buddy allcator is responsible for it. i.e. is it
> done at mm initialization phase or not?
The virtual address coresponding to a physical address for the kernel
module is always the same ..... that doesn't depends on the contexts
...... I think the __va and __pa are well self explainatory or I m not
getting what you can't understand or want to know ..... AFAIK the
virtual mappings of ZONE_NORMAL (__page_tables__) (0MB to 896MB) are
created at the boot time and they are related to the
__page_table_management__ not to the __process_managemet__
> I looked at sources and see in mm/page_alloc.c
>
> #ifdef WANT_PAGE_VIRTUAL
> /* The shift won't overflow because ZONE_NORMAL is
> below 4G. */
> if (!is_highmem_idx(zone))
> set_page_address(page, __va(start_pfn <<
> PAGE_SHIFT));
> #endif
>
> But this assumes WANT_PAGE_VIRTUAL...
>
I don't think on i386 __normally__ that WANT_PAGE_VIRTUAL is defined
in kernel configuration, although this is used to add one more field
to the " struct page " at compile time, so that if the architecture is
slow in multiplication then keeps the virtual address mapping in the
"struct page" ..... here the lines from the include/linux/mm.h
struct page {
................. /*other fields*/ .................
/*
* On machines where all RAM is mapped into kernel address space,
* we can simply calculate the virtual address. On machines with
* highmem some memory is mapped into kernel virtual memory
* dynamically, so we need a place to store that address.
* Note that this field could be 16 bits on x86 ... ;)
*
* Architectures with slow multiplication can define
* WANT_PAGE_VIRTUAL in asm/page.h
*/
#if defined(WANT_PAGE_VIRTUAL)
void *virtual; /* Kernel virtual address (NULL if
not kmapped, ie. highmem) */
#endif /* WANT_PAGE_VIRTUAL */
};
Now I hope it will be more clear then before !!!!!
--
Fawad Lateef
--
Kernelnewbies: Help each other learn about the Linux kernel.
Archive: http://mail.nl.linux.org/kernelnewbies/
FAQ: http://kernelnewbies.org/faq/
