Re: [rfc] optimise unlock_page

[Nick Piggin <npiggin@suse.de>]

On Wed, May 16, 2007 at 06:21:09PM +0100, Hugh Dickins wrote:
> On Sun, 13 May 2007, Nick Piggin wrote:
> > On Fri, May 11, 2007 at 02:15:03PM +0100, Hugh Dickins wrote:
> > 
> > > But again I wonder just what the gain has been, once your double
> > > unmap_mapping_range is factored in.  When I suggested before that
> > > perhaps the double (well, treble including the one in truncate.c)
> > > unmap_mapping_range might solve the problem you set out to solve
> > > (I've lost sight of that!) without pagelock when faulting, you said:
> > > 
> > > > Well aside from being terribly ugly, it means we can still drop
> > > > the dirty bit where we'd otherwise rather not, so I don't think
> > > > we can do that.
> > > 
> > > but that didn't give me enough information to agree or disagree.
> > 
> > Oh, well invalidate wants to be able to skip dirty pages or have the
> > filesystem do something special with them first. Once you have taken
> > the page out of the pagecache but still mapped shared, then blowing
> > it away doesn't actually solve the data loss problem... only makes
> > the window of VM inconsistency smaller.
> 
> Right, I think I see what you mean now, thanks: userspace
> must not for a moment be allowed to write to orphaned pages.

Yep.


> Whereas it's not an issue for the privately COWed pages you added
> the second unmap_mapping_range for: because it's only truncation
> that has to worry about them, so they're heading for SIGBUS anyway.
> 
> Yes, and the page_mapped tests in mm/truncate.c are just racy
> heuristics without the page lock you now put into faulting.

Yes.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/