blob: c02a003fa03a8b9580dbabf60feda4fb41952fa3 [file] [log] [blame]
Hollis Blanchardbbf45ba2008-04-16 23:28:09 -05001Hollis Blanchard <hollisb@us.ibm.com>
215 Apr 2008
3
4Various notes on the implementation of KVM for PowerPC 440:
5
6To enforce isolation, host userspace, guest kernel, and guest userspace all
7run at user privilege level. Only the host kernel runs in supervisor mode.
8Executing privileged instructions in the guest traps into KVM (in the host
9kernel), where we decode and emulate them. Through this technique, unmodified
10440 Linux kernels can be run (slowly) as guests. Future performance work will
11focus on reducing the overhead and frequency of these traps.
12
13The usual code flow is started from userspace invoking an "run" ioctl, which
14causes KVM to switch into guest context. We use IVPR to hijack the host
15interrupt vectors while running the guest, which allows us to direct all
16interrupts to kvmppc_handle_interrupt(). At this point, we could either
17- handle the interrupt completely (e.g. emulate "mtspr SPRG0"), or
18- let the host interrupt handler run (e.g. when the decrementer fires), or
19- return to host userspace (e.g. when the guest performs device MMIO)
20
21Address spaces: We take advantage of the fact that Linux doesn't use the AS=1
22address space (in host or guest), which gives us virtual address space to use
23for guest mappings. While the guest is running, the host kernel remains mapped
24in AS=0, but the guest can only use AS=1 mappings.
25
26TLB entries: The TLB entries covering the host linear mapping remain
27present while running the guest. This reduces the overhead of lightweight
28exits, which are handled by KVM running in the host kernel. We keep three
29copies of the TLB:
30 - guest TLB: contents of the TLB as the guest sees it
31 - shadow TLB: the TLB that is actually in hardware while guest is running
32 - host TLB: to restore TLB state when context switching guest -> host
33When a TLB miss occurs because a mapping was not present in the shadow TLB,
34but was present in the guest TLB, KVM handles the fault without invoking the
35guest. Large guest pages are backed by multiple 4KB shadow pages through this
36mechanism.
37
38IO: MMIO and DCR accesses are emulated by userspace. We use virtio for network
39and block IO, so those drivers must be enabled in the guest. It's possible
40that some qemu device emulation (e.g. e1000 or rtl8139) may also work with
41little effort.