+maintainers and other interested parties On Tue, Jun 27, 2023, Sean Christopherson wrote: > No topic this week, and I need to reclaim some time this week as I will be OOO > all of next week. > > For future topics, a few things on my radar that I am hoping to discuss in the > not-too-distant future, but that need additional work before they're worth > discussing: > > - Coordinating guest_mem() development. I need to post patches, plan is to do > that the week after I get back. > > - Overhauling KVM's gfn_to_pfn() APIs. Need a status update from David S., e.g. > I don't even know if this being actively worked. > > https://lore.kernel.org/all/ZGvUsf7lMkrNDHuE@xxxxxxxxxx > > - KVM + UFFD scalability. We're not yet at the point where we need a synchronous > discussion, but I suspect we'll want a live discussion before merging. > > https://lore.kernel.org/all/20230602161921.208564-1-amoorthy@xxxxxxxxxx > > - Hiding KVM internals from the kernel at large, e.g. moving kvm_host.h into > arch/<arch>/kvm and virt/kvm/, and exporting "internal" KVM symbols if and > only if there are vendor modules. Needs an RFC from us (Google GCE people). Actually, one semi-urgent topic that I'd like to discuss is the LPC KVM Microconference. Specifically, how we want to utilize the allotted three hours. E.g. we can squeeze in ~9 topics if we do 20 minutes per topic, but 20 minutes feels too short for some of the "big" topics like guest_mem(), fine-grained permissions, KVM vs. perf, and pKVM. But I don't want to focus on just 3-4 topics and not leave time to sync on other things. IMO, a mix would work well, e.g. use 2 hours for 3-4 big topics, and then 1 hour for 4-6 small topics. I'm thinking we could use a PUCK session in mid July to get input on what folks generally want the KVM MC to look like so that we can broadcast rough details. And then maybe another session in the September timeframe (after submissions have closed) to go over the "official" schedule before it becomes officially official. Thoughts?
