Re: one question about LRU mechanism

[William Lee Irwin III]

On Wed, Jun 15, 2005 at 01:12:56PM +0800, liyu@WAN wrote:
> 	I am reading memory managment code of kernel 2.6.11.11.
> I found LRU is implement as linked-stack in linux, include two important
> data structure linked-list :
> zone->active_list and zone->inactive_list. when kernel need reclaim some
> pages, it will call function refiil_inactive_list() ultimately to move
> some page from active_list to inactive_list.

The "LRU" bits there don't actually describe the homebrew algorithm.


On Wed, Jun 15, 2005 at 01:12:56PM +0800, liyu@WAN wrote:
> 	It's OK, but I have one question in my mind:
> I found all function that append page to active_list, it just append
> page to head of active_list (use inline function list_add() ), but
> refill_inactive_list() also start scanning from head of active_list, I
> think the better way scan active_list is start from rear of active_list
> and scan though prev member of list_head at reclaim pages. Scanning
> start from head of active_list may make thrashing more possibly, I
> think. and, in my view, "head of active_list" is zone->active_list,
> "rear of active_list" is zone->active_list.prev .
> 	May be, I am failed understand mm? or what's wrong?

I'm looking at 2.6.12-rc6-mm1.

As far as I can tell new active pages are added to the ->next component
of the head of the active list in lru_cache_add_active() and the ->next
component of the head of the inactive list in lru_cache_add(). Then
isolate_lru_pages() dequeues from ->prev of the head of the inactive
list in shrink_cache() and isolate_lru_pages() dequeues from the ->prev
of the active list in refill_inactive_zone().


-- wli
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/