Re: [Xen-devel] [PATCH v2 4/4] x86: fix pinned cache attribute handling

[Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx>]

On Fri, 4 Apr 2014, Jan Beulich wrote:
> >>> On 04.04.14 at 16:30, <stefano.stabellini@xxxxxxxxxxxxx> wrote:
> > On Mon, 31 Mar 2014, Jan Beulich wrote:
> >> >>> On 28.03.14 at 19:00, <stefano.stabellini@xxxxxxxxxxxxx> wrote:
> >> > I miss some context here.
> >> > What is the issue with xc_domain_pin_memory_cacheattr_range and how does
> >> > it affect QEMU (that uses the xc_domain_pin_memory_cacheattr variety)?
> >> 
> >> The issue is that the hypervisor (and hence libxc) interface expect the
> >> passed range to be inclusive, yet the ending page number all the qemus
> >> pass is one past the intended range.
> > 
> > Thanks for the clear explanation.
> > Is this patch what you are looking for?
> > 
> > ---
> > 
> > diff --git a/xen-all.c b/xen-all.c
> > index ba34739..027e7a8 100644
> > --- a/xen-all.c
> > +++ b/xen-all.c
> > @@ -323,7 +323,7 @@ go_physmap:
> >  
> >      xc_domain_pin_memory_cacheattr(xen_xc, xen_domid,
> >                                     start_addr >> TARGET_PAGE_BITS,
> > -                                   (start_addr + size) >> TARGET_PAGE_BITS,
> > +                                   (start_addr + size - 1) >> 
> > TARGET_PAGE_BITS,
> >                                     XEN_DOMCTL_MEM_CACHEATTR_WB);
> >  
> >      snprintf(path, sizeof(path),
> 
> Something along these lines, for all maintained trees. The main thing
> I'm not sure about is whether size can ever be zero, or not page
> aligned - in both cases further care would need to be taken.

size cannot be zero, see:

    if (size <= 0) {
        return -1;
    }

at the beginning of the function.
Currently I don't think that size can be page misaligned. Even if it
could I think that the above would do the right thing.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel