On Tue, 2021-11-02 at 18:19 +0100, Paolo Bonzini wrote: > Yes, I agree. What I am saying is that: > > - the map/unmap dance is not (entirely) about whether to pin the page > - the map/unmap API is not a bad API, just an incomplete implementation > Yep, fair enough. The Xen evtchn series contains what I believe is necessary to make it a complete implementation. But in *this* case it's fairly gratuitous since, as noted, we already *have* a perfectly serviceable mapping. > The GFN _also_ has to be marked dirty. Argh, yes. I forgot to do that. Will fix in v3. Thanks.
Attachment:
smime.p7s
Description: S/MIME cryptographic signature
