On Fri, Sep 28, 2018 at 12:47:08PM +0200, Auger Eric wrote:
> Hi Peter,
Hi, Eric,
[...]
> > +static void vtd_address_space_refresh_all(IntelIOMMUState *s)
> > +{
> > + vtd_address_space_unmap_all(s);
> What about internal context and iotlb caches? Shouldn't they be also
> invalidated at the same time we invalidate external caches?
For vtd_reset(), it's already done in vtd_init().
Though indeed after I tried to cover the two GCMD updates then
possibly we'd better drop the caches there too.
I'll post another version to address the issue.
Thanks for the review,
--
Peter Xu
