Hi guys, After spending 3-4 days adapting KVM user space for ARM and not getting very far Christoffer and I sat down today and went over the project status. The remaining tasks are at least: 1. Understanding QEMU architecture and device emulation 2. Create KVM userspace infrastructure for ARM 3. Write interface between KVM and QEMU for ARM (both kernel and user space code) 4. Finish emulation work 5. Memory architecture The first two tasks turned out to be even more challenging than expected. Just figuring out the basic code structure has taken a lot of time (partly because there are two different KVM userspace implementations by Qumranet and IBM intermixed). We're in dialog with Hollis Blanchard (the author of KVM for PowerPC) regarding a lot of these issues. We estimate that these two tasks will take approximately two weeks. For instance we just figured out that on advice from the QEMU mailing list we have been using the Qumranet version whereas according to Hollis we should have used the IBM version... Task 3 is dependent on 1 and 2 and will take a lot of time considering that we WILL run into numerous unexpected problems when trying to boot an actual kernel and running the binary translation with emulation. Thus we see a big risk that we will not have a running prototype by May 15th. It's better to be realistic and flag for this earlier than later. Jason, how do you feel about this? From our point of view we are already taking the shortest path possible. We will continue our work as planned but we do not anticipate writing code, which uses the KVM module within the above mentioned two weeks. It is our feeling that both Brian and Dave also has plenty of tasks for the upcoming two weeks? Best regards, Andreas & Christoffer
