Re: [PATCH 1/2]: Fix BUG in cancel_dirty_pages on XFS

[Peter Zijlstra <a.p.zijlstra@chello.nl>]

On Thu, 2007-01-25 at 00:43 +1100, Nick Piggin wrote:

> > Have you seen the new launder_page() a_op? called from
> > invalidate_inode_pages2_range()
> 
> It would have been nice to make that one into a more potentially
> useful generic callback.

That can still be done when the need arises, right?

> But why was it introduced, exactly? I can't tell from the code or
> the discussion why NFS couldn't start the IO, and signal the caller
> to wait_on_page_writeback and retry? That seemed to me like the
> convetional fix.

to quote a bit:

On Tue, 19 Dec 2006 18:19:38 -0500
Trond Myklebust <trond.myklebust@fys.uio.no> wrote:

>     NFS: Fix race in nfs_release_page()
>     
>     invalidate_inode_pages2() may set the dirty bit on a page owing to the call
>     to unmap_mapping_range() after the page was locked. In order to fix this,
>     NFS has hooked the releasepage() method. This, however leads to deadlocks
>     in other parts of the VM.

and:

> > Now, arguably the VM shouldn't be calling try_to_release_page() with
> > __GFP_FS when it's holding a lock on a page.
> > 
> > But otoh, NFS should never be running lock_page() within nfs_release_page()
> > against the page which was passed into nfs_release_page().  It'll deadlock
> > for sure.
> 
> The reason why it is happening is that the last dirty page from that
> inode gets cleaned, resulting in a call to dput().