On Tue, Nov 23, 2021 at 10:06:02AM +0100, Paolo Bonzini wrote: > On 11/19/21 16:39, David Hildenbrand wrote: > > > If qmeu can put all the guest memory in a memfd and not map it, then > > > I'd also like to see that the IOMMU can use this interface too so we > > > can have VFIO working in this configuration. > > > > In QEMU we usually want to (and must) be able to access guest memory > > from user space, with the current design we wouldn't even be able to > > temporarily mmap it -- which makes sense for encrypted memory only. The > > corner case really is encrypted memory. So I don't think we'll see a > > broad use of this feature outside of encrypted VMs in QEMU. I might be > > wrong, most probably I am:) > > It's not _that_ crazy an idea, but it's going to be some work to teach KVM > that it has to kmap/kunmap around all memory accesses. > > I think it's great that memfd hooks are usable by more than one subsystem, > OTOH it's fair that whoever needs it does the work---and VFIO does not need > it for confidential VMs, yet, so it should be fine for now to have a single > user. > > On the other hand, as I commented already, the lack of locking in the > register/unregister functions has to be fixed even with a single user. > Another thing we can do already is change the guest_ops/guest_mem_ops to > something like memfd_falloc_notifier_ops/memfd_pfn_ops, and the > register/unregister functions to memfd_register/unregister_falloc_notifier. I'm satisified with this naming ;) > > Chao, can you also put this under a new CONFIG such as "bool MEMFD_OPS", and > select it from KVM? Yes, reasonable. > > Thanks, > > Paolo
