Re: [RFC PATCH V1 08/12] xen/arm: Invalidate qemu mapcache on XENMEM_decrease_reservation

[Oleksandr <olekstysh@xxxxxxxxx>]

On 06.08.20 19:33, Jan Beulich wrote:

Hi Jan.

> On 06.08.2020 16:28, Oleksandr wrote:
> > On 06.08.20 14:50, Jan Beulich wrote:
> >
> > Hi Jan
> >
> > > > > On 03.08.2020 20:21, Oleksandr Tyshchenko wrote:
> > > > > > --- a/xen/common/memory.c
> > > > > > +++ b/xen/common/memory.c
> > > > > > @@ -1652,6 +1652,12 @@ long do_memory_op(unsigned long cmd, 
> > > > > > XEN_GUEST_HANDLE_PARAM(void) arg)
> > > > > >             break;
> > > > > >         }
> > > > > >     
> > > > > > +    /* x86 already sets the flag in hvm_memory_op() */
> > > > > > +#if defined(CONFIG_ARM64) && defined(CONFIG_IOREQ_SERVER)
> > > > > > +    if ( op == XENMEM_decrease_reservation )
> > > > > > +        curr_d->arch.hvm.qemu_mapcache_invalidate = true;
> > > > > > +#endif
> > > > > Doesn't the comment already indicate a route towards an approach
> > > > > not requiring to alter common code?
> > > > Given that IOREQ is now moved under common/, I think it would make sense
> > > > to have this set in common code as well for all the architecture.
> > > >
> > > > IOW, I would suggest to drop the #ifdef CONFIG_ARM64. In addition, we
> > > > may want to introduce an helper to check if a domain is using ioreq.
> > > Of course, with the (part of the) conditional dropped and the struct
> > > field moved out of the arch sub-struct, this is fine to live here.
> > ok.
> >
> >
> > I suspect this should *also* live in compat_memory_op(). Please confirm
> > whether my understanding correct.
> Doesn't compat_memory_op() simply call here, so will have the flag set
> as needed?
Indeed, sorry for the noise.


--
Regards,

Oleksandr Tyshchenko