RE: [PATCH 3/5] libxl / iommu / domctl: introduce XEN_DOMCTL_IOMMU_SET_ALLOCATION...

[Paul Durrant <xadimgnik@xxxxxxxxx>]

> -----Original Message-----
> From: Julien Grall <julien@xxxxxxx>
> Sent: 16 October 2020 16:55
> To: Paul Durrant <paul@xxxxxxx>; xen-devel@xxxxxxxxxxxxxxxxxxxx
> Cc: Paul Durrant <pdurrant@xxxxxxxxxx>; Ian Jackson <iwj@xxxxxxxxxxxxxx>; Wei 
> Liu <wl@xxxxxxx>; Andrew
> Cooper <andrew.cooper3@xxxxxxxxxx>; George Dunlap <george.dunlap@xxxxxxxxxx>; 
> Jan Beulich
> <jbeulich@xxxxxxxx>; Stefano Stabellini <sstabellini@xxxxxxxxxx>; Anthony 
> PERARD
> <anthony.perard@xxxxxxxxxx>; Volodymyr Babchuk <Volodymyr_Babchuk@xxxxxxxx>; 
> Roger Pau Monné
> <roger.pau@xxxxxxxxxx>
> Subject: Re: [PATCH 3/5] libxl / iommu / domctl: introduce 
> XEN_DOMCTL_IOMMU_SET_ALLOCATION...
> 
> Hi Paul,
> 
> On 05/10/2020 10:49, Paul Durrant wrote:
> > From: Paul Durrant <pdurrant@xxxxxxxxxx>
> >
> > ... sub-operation of XEN_DOMCTL_iommu_ctl.
> >
> > This patch adds a new sub-operation into the domctl. The code in iommu_ctl()
> > is extended to call a new arch-specific iommu_set_allocation() function 
> > which
> > will be called with the IOMMU page-table overhead (in 4k pages) in response
> 
> Why 4KB? Wouldn't it be better to use the hypervisor page size instead?
> 

I think I'll follow the shadow/hap code more closely and just pass a value in 
MB, then any issue with page size is left inside Xen.

> > diff --git a/xen/include/public/domctl.h b/xen/include/public/domctl.h
> > index 75e855625a..6402678838 100644
> > --- a/xen/include/public/domctl.h
> > +++ b/xen/include/public/domctl.h
> > @@ -1138,8 +1138,16 @@ struct xen_domctl_vuart_op {
> >
> >   #define XEN_DOMCTL_IOMMU_INVALID 0
> >
> > +#define XEN_DOMCTL_IOMMU_SET_ALLOCATION 1
> > +struct xen_domctl_iommu_set_allocation {
> > +    uint32_t nr_pages;
> 
> Shouldn't this be a 64-bit value?

If I pass the value in MB then 32-bits will cover it, I think. I do need to add 
padding though.

  Paul

> 
> Cheers,
> 
> --
> Julien Grall