Re: [PATCH] double flush_page_to_ram

[David S. Miller]

   From: Hugh Dickins <[e-mail]>
   Date: Thu, 9 May 2002 14:18:49 +0100 (BST)

   On Thu, 9 May 2002, David S. Miller wrote:
   > Wrong, consider the case where we do early COW in do_no_page, you miss
   > a flush on the new-new page.
   
   Of course we do, and then we don't map it into user address space;
   if it ever gets mapped into user address space later, do_no_page
   does the flush_page_to_ram then.

You miss the fact that if we do an early COW and another process
recently WROTE into that page via a shared MMAP, we will potentially
copy old data into the COW page we use for the current process.

Your changes are wrong and will cause corruption.