Thank you Rik, u saved me from getting one more white hair.....:-).
On Sat, Jun 14, 2008 at 11:30 AM, Rik van Riel <[EMAIL PROTECTED]> wrote:
> On Sat, 14 Jun 2008 10:51:19 +0800
> "Peter Teoh" <[EMAIL PROTECTED]> wrote:
>
>> first, it should attempt to allocate space from physical mem...and it
>> uses max_pfn as the mark to guide it.
>
> No.
>
>> On the other hand, because i cannot find the link between max_pfn and
>> __get_free_pages() - so question is how does __get_free_pages() know
>> where the max pfn it can used for memory allocation?
>
> You read correctly, __get_free_pages() has absolutely nothing to
> do with max_pfn.
>
>> next, it should then update the pagetable entries - right?\
>
> No, __get_free_pages() never manipulates page tables.
>
>> essentially...it should be marking the pagetable entries for the block
>> of memory as "notfree" - right? what is the function used? thanks.
>
> This info is not kept in page tables, but in the array
> struct page mem_map[].
>
> Take a look at the functions that __get_free_pages() calls:
>
> alloc_pages()
> -> alloc_pages_node()
> -> __alloc_pages()
> -> __alloc_pages_internal()
> -> get_page_from_freelist()
> -> buffered_rmqueue()
> -> rmqueue_bulk()
> -> __rmqueue()
> -> __rmqueue_smallest()
>
> As you can see, each zone has a number of free lists, for each
> allocation order:
>
> /* Find a page of the appropriate size in the preferred list */
> for (current_order = order; current_order < MAX_ORDER;
> ++current_order) {
> area = &(zone->free_area[current_order]);
> if (list_empty(&area->free_list[migratetype]))
> continue;
>
> page = list_entry(area->free_list[migratetype].next,
> struct page, lru);
Yes....as I read the code now I get better understanding. Thanks for the lead.
But now I have to shift the entire question to another area again.
Original intent was understanding how the pagetable got updated. So,
as u implied, it is not done at mem allocation time, but when the
zone-based free list is constructed - for each order - right?
I have traced a few functions that are involved when constructing the
area->free_list (quoted from your extract above):
free_area_init()
free_area_init_node()
free_area_init_core()
init_currently_empty_zones(), memmap_init()
zone_init_free_lists()
And the one specifically involved in pagetable update is memmap_init():
void __meminit memmap_init_zone(unsigned long size, int nid, unsigned long zone,
unsigned long start_pfn, enum memmap_context context)
{
struct page *page;
unsigned long end_pfn = start_pfn + size;
unsigned long pfn;
struct zone *z;
z = &NODE_DATA(nid)->node_zones[zone];
for (pfn = start_pfn; pfn < end_pfn; pfn++) {
But this node_zones-->is not the pagetable? How is it linked to the pagetable?
thanks.
--
Regards,
Peter Teoh
--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to [EMAIL PROTECTED]
Please read the FAQ at http://kernelnewbies.org/FAQ
