Re: unuse_pte: set pte dirty if the page is dirty

[Andrew Morton <akpm@osdl.org>]

Christoph Lameter <clameter@engr.sgi.com> wrote:
>
> On Mon, 27 Feb 2006, Andrew Morton wrote:
> 
> > Are we sure this is race-free?  Say, someone is in the process of cleaning
> > the page?  munmap, conceivably swapout?  We end up with a dirty pte
> > pointing at a now-clean page.  The page will later become dirty again.  Is
> > that a problem?  It would generate a surprise if the vma had ben set
> > read-only in the interim, for example.
> 
> munmap sets the dirty bit in pages rather than clearing the dirty bits.
> 
> If we would set a dirty bit in a pte pointing to a now clean page then 
> unmapping (or the swaper) will mark the page dirty again and its going to 
> be rewritten again.

Precisely.

And will that crash the kernel, corrupt swapspace, or any other such
exciting things?

There are cases (eg, mprotect) in which a subsequent page-dirtying is
"impossible".  Only we've now gone and made it possible.  The worst part of
it is that we're made it possible in exceedingly rare circumstances.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>