Re: [e-mail]

[Andrew Morton]

Ulrich Weigand wrote:
> 
> Momchil Velikov wrote:
> 
> +int move_from_swap_cache(struct page *page, unsigned long index,
> +               struct address_space *mapping)
> +{
> +       void **pslot;
> +       int err;
> +
> +       if (!PageLocked(page))
> +               BUG();
> +
> +       spin_lock(&swapper_space.i_shared_lock);
> +       spin_lock(&mapping->i_shared_lock);
> +
> +       err = radix_tree_reserve(&mapping->page_tree, index, &pslot);
> +       if (!err) {
> +               swp_entry_t entry;
> +
> +               block_flushpage(page, 0);

block_flushpage inside spinlock is a no-no.  It does I/O,
and sleeps.

-