master
rst 68 lines 2.59 KB
Raw
1 ==============
2 UEFI variables
3 ==============
4
5 Guest UEFI variable management
6 ==============================
7
8 The traditional approach for UEFI Variable storage in qemu guests is
9 to work as close as possible to physical hardware. That means
10 providing pflash as storage and leaving the management of variables
11 and flash to the guest.
12
13 Secure boot support comes with the requirement that the UEFI variable
14 storage must be protected against direct access by the OS. All update
15 requests must pass the sanity checks. (Parts of) the firmware must
16 run with a higher privilege level than the OS so this can be enforced
17 by the firmware. On x86 this has been implemented using System
18 Management Mode (SMM) in qemu and kvm, which again is the same
19 approach taken by physical hardware. Only privileged code running in
20 SMM mode is allowed to access flash storage.
21
22 Communication with the firmware code running in SMM mode works by
23 serializing the requests to a shared buffer, then trapping into SMM
24 mode via SMI. The SMM code processes the request, stores the reply in
25 the same buffer and returns.
26
27 Host UEFI variable service
28 ==========================
29
30 Instead of running the privileged code inside the guest we can run it
31 on the host. The serialization protocol can be reused. The
32 communication with the host uses a virtual device, which essentially
33 configures the shared buffer location and size, and traps to the host
34 to process the requests.
35
36 The ``uefi-vars`` device implements the UEFI virtual device. It comes
37 in ``uefi-vars-x64`` and ``uefi-vars-sysbus`` flavours. The device
38 reimplements the handlers needed, specifically
39 ``EfiSmmVariableProtocol`` and ``VarCheckPolicyLibMmiHandler``. It
40 also consumes events (``EfiEndOfDxeEventGroup``,
41 ``EfiEventReadyToBoot`` and ``EfiEventExitBootServices``).
42
43 The advantage of the approach is that we do not need a special
44 privilege level for the firmware to protect itself, i.e. it does not
45 depend on SMM emulation on x64, which allows the removal of a bunch of
46 complex code for SMM emulation from the linux kernel
47 (CONFIG_KVM_SMM=n). It also allows support for secure boot on arm
48 without implementing secure world (el3) emulation in kvm.
49
50 Of course there are also downsides. The added device increases the
51 attack surface of the host, and we are adding some code duplication
52 because we have to reimplement some edk2 functionality in qemu.
53
54 usage on x86_64
55 ---------------
56
57 .. code::
58
59 qemu-system-x86_64 \
60 -device uefi-vars-x64,jsonfile=/path/to/vars.json
61
62 usage on aarch64
63 ----------------
64
65 .. code::
66
67 qemu-system-aarch64 -M virt \
68 -device uefi-vars-sysbus,jsonfile=/path/to/vars.json