1.. SPDX-License-Identifier: GPL-2.0
2
3==============
4Nitro Enclaves
5==============
6
7Overview
8========
9
10Nitro Enclaves (NE) is a new Amazon Elastic Compute Cloud (EC2) capability
11that allows customers to carve out isolated compute environments within EC2
12instances [1].
13
14For example, an application that processes sensitive data and runs in a VM,
15can be separated from other applications running in the same VM. This
16application then runs in a separate VM than the primary VM, namely an enclave.
17
18An enclave runs alongside the VM that spawned it. This setup matches low latency
19applications needs. The resources that are allocated for the enclave, such as
20memory and CPUs, are carved out of the primary VM. Each enclave is mapped to a
21process running in the primary VM, that communicates with the NE driver via an
22ioctl interface.
23
24In this sense, there are two components:
25
261. An enclave abstraction process - a user space process running in the primary
27VM guest that uses the provided ioctl interface of the NE driver to spawn an
28enclave VM (that's 2 below).
29
30There is a NE emulated PCI device exposed to the primary VM. The driver for this
31new PCI device is included in the NE driver.
32
33The ioctl logic is mapped to PCI device commands e.g. the NE_START_ENCLAVE ioctl
34maps to an enclave start PCI command. The PCI device commands are then
35translated into  actions taken on the hypervisor side; that's the Nitro
36hypervisor running on the host where the primary VM is running. The Nitro
37hypervisor is based on core KVM technology.
38
392. The enclave itself - a VM running on the same host as the primary VM that
40spawned it. Memory and CPUs are carved out of the primary VM and are dedicated
41for the enclave VM. An enclave does not have persistent storage attached.
42
43The memory regions carved out of the primary VM and given to an enclave need to
44be aligned 2 MiB / 1 GiB physically contiguous memory regions (or multiple of
45this size e.g. 8 MiB). The memory can be allocated e.g. by using hugetlbfs from
46user space [2][3]. The memory size for an enclave needs to be at least 64 MiB.
47The enclave memory and CPUs need to be from the same NUMA node.
48
49An enclave runs on dedicated cores. CPU 0 and its CPU siblings need to remain
50available for the primary VM. A CPU pool has to be set for NE purposes by an
51user with admin capability. See the cpu list section from the kernel
52documentation [4] for how a CPU pool format looks.
53
54An enclave communicates with the primary VM via a local communication channel,
55using virtio-vsock [5]. The primary VM has virtio-pci vsock emulated device,
56while the enclave VM has a virtio-mmio vsock emulated device. The vsock device
57uses eventfd for signaling. The enclave VM sees the usual interfaces - local
58APIC and IOAPIC - to get interrupts from virtio-vsock device. The virtio-mmio
59device is placed in memory below the typical 4 GiB.
60
61The application that runs in the enclave needs to be packaged in an enclave
62image together with the OS ( e.g. kernel, ramdisk, init ) that will run in the
63enclave VM. The enclave VM has its own kernel and follows the standard Linux
64boot protocol [6].
65
66The kernel bzImage, the kernel command line, the ramdisk(s) are part of the
67Enclave Image Format (EIF); plus an EIF header including metadata such as magic
68number, eif version, image size and CRC.
69
70Hash values are computed for the entire enclave image (EIF), the kernel and
71ramdisk(s). That's used, for example, to check that the enclave image that is
72loaded in the enclave VM is the one that was intended to be run.
73
74These crypto measurements are included in a signed attestation document
75generated by the Nitro Hypervisor and further used to prove the identity of the
76enclave; KMS is an example of service that NE is integrated with and that checks
77the attestation doc.
78
79The enclave image (EIF) is loaded in the enclave memory at offset 8 MiB. The
80init process in the enclave connects to the vsock CID of the primary VM and a
81predefined port - 9000 - to send a heartbeat value - 0xb7. This mechanism is
82used to check in the primary VM that the enclave has booted. The CID of the
83primary VM is 3.
84
85If the enclave VM crashes or gracefully exits, an interrupt event is received by
86the NE driver. This event is sent further to the user space enclave process
87running in the primary VM via a poll notification mechanism. Then the user space
88enclave process can exit.
89
90[1] https://aws.amazon.com/ec2/nitro/nitro-enclaves/
91[2] https://www.kernel.org/doc/html/latest/admin-guide/mm/hugetlbpage.html
92[3] https://lwn.net/Articles/807108/
93[4] https://www.kernel.org/doc/html/latest/admin-guide/kernel-parameters.html
94[5] https://man7.org/linux/man-pages/man7/vsock.7.html
95[6] https://www.kernel.org/doc/html/latest/x86/boot.html
96