On Wed, Apr 29, 2020 at 10:40 AM Paolo Bonzini <pbonzini@xxxxxxxxxx> wrote: > > On 29/04/20 19:28, Andy Lutomirski wrote: > > This seems functional, but I'm wondering if it could a bit simpler and > > more efficient if the data structure was a normal descriptor ring with > > the same number slots as whatever the maximum number of waiting pages > > is. Then there would never need to be any notification from the guest > > back to the host, since there would always be room for a notification. > > No, it would be much more complicated code for a slow path which is > already order of magnitudes slower than a vmexit. It would also use > much more memory. Fair enough. > > > It might be even better if a single unified data structure was used > > for both notifications. > > That's a very bad idea since one is synchronous and one is asynchronous. > Part of the proposal we agreed upon was to keep "page not ready" > synchronous while making "page ready" an interrupt. The data structure > for "page not ready" will be #VE. > #VE on SVM will be interesting, to say the least, and I think that a solution that is VMX specific doesn't make much sense. #VE also has unpleasant issues involving the contexts in which it can occur. You will have quite a hard time convincing me to ack the addition of a #VE entry handler for this. I think a brand new vector is the right solution.
