Re: [Xen-devel] [PATCH V3] tools/libxc, xen/x86: Added xc_set_mem_access_multi()

[Julien Grall <julien.grall@xxxxxxx>]

On 06/09/16 12:05, George Dunlap wrote:
> On 06/09/16 11:51, Julien Grall wrote:
> > On 06/09/16 11:45, Razvan Cojocaru wrote:
> > > Hello,
> > >
> > > > On 06/09/16 11:00, Razvan Cojocaru wrote:
> > > > > diff --git a/xen/arch/arm/p2m.c b/xen/arch/arm/p2m.c
> > > > > index b648a9d..e65a9b8 100644
> > > > > --- a/xen/arch/arm/p2m.c
> > > > > +++ b/xen/arch/arm/p2m.c
> > > > > @@ -1836,6 +1836,15 @@ long p2m_set_mem_access(struct domain *d, gfn_t
> > > > > gfn, uint32_t nr,
> > > > >      return 0;
> > > > >  }
> > > > >
> > > > > +long p2m_set_mem_access_multi(struct domain *d,
> > > > > +                              const XEN_GUEST_HANDLE(const_uint64)
> > > > > pfn_list,
> > > > > +                              const XEN_GUEST_HANDLE(const_uint8)
> > > > > access_list,
> > > > > +                              uint32_t nr, uint32_t start, uint32_t
> > > > > mask,
> > > > > +                              unsigned int altp2m_idx)
> > > > > +{
> > > > > +    return -ENOTSUP;
> > > > Why didn't you implement this function for ARM?
> > > Because unfortunately I don't have an ARM setup to test it on and I
> > > thought it would be unfair to publish the patch with untestable ARM
> > > support.
> > So what's the plan? Who will implement the ARM solution?
> >
> > I don't think there is a technical challenge to implement the ARM one.
> Are we going to require that all new functionality be implemented both
> on x86 and on ARM?  That seems like a bit of a lot to ask of someone
> who's not going to use it (and as Razvan points out, can't even test
> it).  The mem_access functionality is still fairly technical -- anyone
> who decides they want to use it and runs across the same problem Razvan
> did should have no trouble implementing this hypercall for ARM when they
> need it.
I am not saying we should every time implement the ARM version when the 
x86 version is added and vice-versa.
However, I don't think this is a lot to ask when an hypercall benefits 
both architecture. Note that nobody would have complained that the code 
was not tested on ARM if the hypercall was implemented in the common 
code. Even if it could break the platform...
Regards,

--
Julien Grall

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