Re: [PATCH] anobjrmap 2/6 mapping

[Andrew Morton]

Hugh Dickins <[e-mail]> wrote:
>
> It is likely that I've screwed up on the "Morton pages", those
> ext3 journal pages locked at truncate time which then turn into
> fish with wings: please check them out, I never manage to wrap
> my head around them.  Certainly don't want a page using private
> for both bufferheads and swp_entry_t.

It goes BUG in try_to_free_buffers().

We really should fix this up for other reasons, probably by making ext3's
per-page truncate operations wait on commit, and be more aggressive about
pulling the page's buffers off the transaction at truncate time.

The same thing _could_ happen with other filesystems; not too sure about
that.

Still.  I suggest you look at freeing up page->list from anon/swapcache
pages.  It really doesn't do much.

Meanwhile, I backed out that bit - I don't actually see where the failure is
anyway.  The page is page_mapped(), !PageAnon and ->mapping == NULL.