Re: 2.4.18(19) swapcache oops

[Andrew Morton <akpm@zip.com.au>]

Andrew Morton wrote:
> 
> ...
> --- 2.4.19/mm/swap.c~lru-race   Thu Aug 15 13:03:48 2002
> +++ 2.4.19-akpm/mm/swap.c       Thu Aug 15 13:04:19 2002
> @@ -57,9 +57,10 @@ void activate_page(struct page * page)
>   */
>  void lru_cache_add(struct page * page)
>  {
> -       if (!TestSetPageLRU(page)) {
> +       if (!PageLRU(page)) {
>                 spin_lock(&pagemap_lru_lock);
> -               add_page_to_inactive_list(page);
> +               if (!TestSetPageLRU(page))
> +                       add_page_to_inactive_list(page);
>                 spin_unlock(&pagemap_lru_lock);
>         }
>  }

Seems that I fixed this in 2.5.32.  That set_bit outside
the lock gave me the willes, and I couldn't put my finger on why.
Never occurred to me that the page could be found via pagecache
lookup in this manner.

In 2.5, it is effectively:

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

which is what should be tested in 2.4.  It's stricter, and significantly
faster.