Re: [RFC] [PATCH] Include LRU in page count

[Andrew Morton]

Christian Ehrhardt wrote:
> 
> ...
> Solution: Change the PageLRU bit inside the lock. Looks like
> lru_cache_add is the only place that doesn't hold the lru lock to
> change the LRU flag and it's probably not a good idea even without
> the patch.
> 

2.4.20-pre already does that:

void lru_cache_add(struct page * page)
{
        if (!PageLRU(page)) {
                spin_lock(&pagemap_lru_lock);
                if (!TestSetPageLRU(page))
                        add_page_to_inactive_list(page);
                spin_unlock(&pagemap_lru_lock);
        }
}

there was an oopsing race with activate_page()...