Re: Nick's core remove PageReserved broke vmware...

["Michael S. Tsirkin" <mst@mellanox.co.il>]

Quoting Gleb Natapov <gleb@minantech.com>:
> Subject: Re: Nick's core remove PageReserved broke vmware...
> 
> On Tue, Nov 08, 2005 at 11:34:07PM +0200, Michael S. Tsirkin wrote:
> > Index: linux-2.6.14-dontcopy/mm/madvise.c
> > ===================================================================
> > --- linux-2.6.14-dontcopy.orig/mm/madvise.c	2005-10-28
> 02:02:08.000000000 +0200
> > +++ linux-2.6.14-dontcopy/mm/madvise.c	2005-11-08
> 23:28:56.000000000 +0200
> > @@ -31,6 +31,12 @@ static long madvise_behavior(struct vm_a
> >  	case MADV_RANDOM:
> >  		new_flags |= VM_RAND_READ;
> >  		break;
> > +	case MADV_DONTCOPY:
> > +		new_flags |= VM_UDONTCOPY;
> > +		break;
> > +	case MADV_DOCOPY:
> > +		new_flags &= ~VM_UDONTCOPY;
> > +		break;
> >  	default:
> >  		break;
> >  	}
> I think you are removing VM_RAND_READ/VM_SEQ_READ here.

True, we need something like

        switch (behavior) {
        case MADV_SEQUENTIAL:
                new_flags = (vma->vm_flags & ~VM_READHINTMASK) | VM_SEQ_READ;
                break;
        case MADV_RANDOM:
                new_flags = (vma->vm_flags & ~VM_READHINTMASK) | VM_RAND_READ;
                break;
        case MADV_DONTCOPY:
                new_flags |= VM_UDONTCOPY;
                break;
        case MADV_DOCOPY:
                new_flags &= ~VM_UDONTCOPY;
                break;
        default:
                break;
        }

Thanks!

> Also perhapse we should skip VM_SHARED VMAs?

Why?

> --
> 			Gleb.
> 

-- 
MST