Re: [RFC PATCH] mm: Replace ->launder_folio() with flush and wait

[David Howells <dhowells@redhat.com>]

Matthew Wilcox <willy@infradead.org> wrote:

> commit e3db7691e9f3dff3289f64e3d98583e28afe03db
> Author: Trond Myklebust <Trond.Myklebust@netapp.com>
> Date:   Wed Jan 10 23:15:39 2007 -0800
> 
>     [PATCH] NFS: Fix race in nfs_release_page()
>... 
>       invalidate_inode_pages2() may find the dirty bit has been set on a page
>       owing to the fact that the page may still be mapped after it was locked.
>       Only after the call to unmap_mapping_range() are we sure that the page
>       can no longer be dirtied.

Is that last sentence even true?  It evicts folios from the TLB and/or
pagetables, but it doesn't actually trim any mmap made - in which case,
userspace is perfectly at liberty to regenerate and dirty the folio the moment
the folio is removed from the page tables.  Otherwise DIO-to/from-mmap will
deadlock.

> but my belief is that we should be able to get rid of it.

I think you're probably correct.  The best we can do, I think, is to preface
any call to invalidate_inode_pages2() with a flush-and-wait over the same
range.

David