Re: [PATCH RFC 4/4] mm: guest_memfd: Add ability for mmap'ing pages

[Elliot Berman <quic_eberman@quicinc.com>]

On Tue, Aug 06, 2024 at 03:51:22PM +0200, David Hildenbrand wrote:
> > -	if (gmem_flags & GUEST_MEMFD_FLAG_NO_DIRECT_MAP) {
> > +	if (!ops->accessible && (gmem_flags & GUEST_MEMFD_FLAG_NO_DIRECT_MAP)) {
> >   		r = guest_memfd_folio_private(folio);
> >   		if (r)
> >   			goto out_err;
> > @@ -107,6 +109,82 @@ struct folio *guest_memfd_grab_folio(struct file *file, pgoff_t index, u32 flags
> >   }
> >   EXPORT_SYMBOL_GPL(guest_memfd_grab_folio);
> > +int guest_memfd_make_inaccessible(struct file *file, struct folio *folio)
> > +{
> > +	unsigned long gmem_flags = (unsigned long)file->private_data;
> > +	unsigned long i;
> > +	int r;
> > +
> > +	unmap_mapping_folio(folio);
> > +
> > +	/**
> > +	 * We can't use the refcount. It might be elevated due to
> > +	 * guest/vcpu trying to access same folio as another vcpu
> > +	 * or because userspace is trying to access folio for same reason
> 
> As discussed, that's insufficient. We really have to drive the refcount to 1
> -- the single reference we expect.
> 
> What is the exact problem you are running into here? Who can just grab a
> reference and maybe do nasty things with it?
> 

Right, I remember we had discussed it. The problem I faced was if 2
vcpus fault on same page, they would race to look up the folio in
filemap, increment refcount, then try to lock the folio. One of the
vcpus wins the lock, while the other waits. The vcpu that gets the
lock vcpu will see the elevated refcount.

I was in middle of writing an explanation why I think this is best
approach and realized I think it should be possible to do
shared->private conversion and actually have single reference. There
would be some cost to walk through the allocated folios and convert them
to private before any vcpu runs. The approach I had gone with was to
do conversions as late as possible.

Thanks,
Elliot