Hello, (Note: I had sent this mail in linux-pci and haven't gotten a response, so trying KVM) I am working on a project in which we are designing the system with several Guest VMs and at least one privileged VM other than the Base Linux. So, the hierarchy looks like as follows Base Linux ----> Priv_VM ----> G1, G2, ... Gn (Guest Vms) We are trying to implement a few HA mechanisms like 1. Protecting the Base Linux by trying to hand over the entire PCIe topology to the Priv_VM and let it do the PCIe pass-through of selected functions to Guest Vms G1, G2 ... I guess this requires two stage hand off and wondering it this kind of thing is doable. What is the general outline to accomplish this. 2. The Priv_VM needs to hand off a specific Function to a Guest G1. We are thinking that it will be better to hand off the device as well as the downstream port to which the device is connected. This will enable the guest VM to add port based hot plug drivers and AER drivers specific to that VM on that downstream port and handle hot plug and other error events? Is this a correct architecture? Any ideas on what kind of issues we will have to deal with here? 3. Often we have to take a PCIe device from Guest VM and give it to other guest Vms. For example, Priv_VM gives the device 0000:06:00.1 to VM G1 and then later (on a certain condition ) grabs the device and hands it over to another VM G2. - We want to ensure that this is doable.? (How do we go about doing this). - We also want the device to be intact as much as possible , ie the device should not be reset. (rewriting the config space is not a concern). Any ideas on how we can get some of these done?. or is this kind of doable direction. pals -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
