Bug in mm.h?

Bug in mm.h?

[Leonid Podolny]

I fail to understand whether the following is a bug. From what I see, if
the page is reserved, page->count is not decreased. The order of the
conditions should be reversed.

>From mm.h:

static inline void put_page(struct page *page)
{
        if (!PageReserved(page) && put_page_testzero(page))
                __page_cache_release(page);
}

static inline void get_page(struct page *page)
{
        atomic_inc(&page->_count);
}

#define put_page_testzero(p)                            \
        ({                                              \
                BUG_ON(page_count(p) == 0);             \
                atomic_add_negative(-1, &(p)->_count);  \
        })