Re: truncate_list_pages() BUG and confusion

[Dave Hansen]

Martin J. Bligh wrote:
> Where did truncate_list_pages unlock it? It doesn't do that
> until after it's called truncate_complete_page, which (indirectly)
> frees the page ... ?

Ahh, put_page_testzero
#define put_page_testzero(p)    atomic_dec_and_test(&(p)->count)

It will only try to free if the page count is zero.  The caller makes 
sure that this doesn't happen.

if (!PageReserved(page) && put_page_testzero(page)) {
	if (PageLRU(page))
         	lru_cache_del(page);
	__free_pages_ok(page, 0);
}

-- 
Dave Hansen
[e-mail]