1.. SPDX-License-Identifier: GPL-2.0
2
3===============================
4Software Guard eXtensions (SGX)
5===============================
6
7Overview
8========
9
10Software Guard eXtensions (SGX) hardware enables for user space applications
11to set aside private memory regions of code and data:
12
13* Privileged (ring-0) ENCLS functions orchestrate the construction of the
14  regions.
15* Unprivileged (ring-3) ENCLU functions allow an application to enter and
16  execute inside the regions.
17
18These memory regions are called enclaves. An enclave can be only entered at a
19fixed set of entry points. Each entry point can hold a single hardware thread
20at a time.  While the enclave is loaded from a regular binary file by using
21ENCLS functions, only the threads inside the enclave can access its memory. The
22region is denied from outside access by the CPU, and encrypted before it leaves
23from LLC.
24
25The support can be determined by
26
27	``grep sgx /proc/cpuinfo``
28
29SGX must both be supported in the processor and enabled by the BIOS.  If SGX
30appears to be unsupported on a system which has hardware support, ensure
31support is enabled in the BIOS.  If a BIOS presents a choice between "Enabled"
32and "Software Enabled" modes for SGX, choose "Enabled".
33
34Enclave Page Cache
35==================
36
37SGX utilizes an *Enclave Page Cache (EPC)* to store pages that are associated
38with an enclave. It is contained in a BIOS-reserved region of physical memory.
39Unlike pages used for regular memory, pages can only be accessed from outside of
40the enclave during enclave construction with special, limited SGX instructions.
41
42Only a CPU executing inside an enclave can directly access enclave memory.
43However, a CPU executing inside an enclave may access normal memory outside the
44enclave.
45
46The kernel manages enclave memory similar to how it treats device memory.
47
48Enclave Page Types
49------------------
50
51**SGX Enclave Control Structure (SECS)**
52   Enclave's address range, attributes and other global data are defined
53   by this structure.
54
55**Regular (REG)**
56   Regular EPC pages contain the code and data of an enclave.
57
58**Thread Control Structure (TCS)**
59   Thread Control Structure pages define the entry points to an enclave and
60   track the execution state of an enclave thread.
61
62**Version Array (VA)**
63   Version Array pages contain 512 slots, each of which can contain a version
64   number for a page evicted from the EPC.
65
66Enclave Page Cache Map
67----------------------
68
69The processor tracks EPC pages in a hardware metadata structure called the
70*Enclave Page Cache Map (EPCM)*.  The EPCM contains an entry for each EPC page
71which describes the owning enclave, access rights and page type among the other
72things.
73
74EPCM permissions are separate from the normal page tables.  This prevents the
75kernel from, for instance, allowing writes to data which an enclave wishes to
76remain read-only.  EPCM permissions may only impose additional restrictions on
77top of normal x86 page permissions.
78
79For all intents and purposes, the SGX architecture allows the processor to
80invalidate all EPCM entries at will.  This requires that software be prepared to
81handle an EPCM fault at any time.  In practice, this can happen on events like
82power transitions when the ephemeral key that encrypts enclave memory is lost.
83
84Application interface
85=====================
86
87Enclave build functions
88-----------------------
89
90In addition to the traditional compiler and linker build process, SGX has a
91separate enclave “build” process.  Enclaves must be built before they can be
92executed (entered). The first step in building an enclave is opening the
93**/dev/sgx_enclave** device.  Since enclave memory is protected from direct
94access, special privileged instructions are then used to copy data into enclave
95pages and establish enclave page permissions.
96
97.. kernel-doc:: arch/x86/kernel/cpu/sgx/ioctl.c
98   :functions: sgx_ioc_enclave_create
99               sgx_ioc_enclave_add_pages
100               sgx_ioc_enclave_init
101               sgx_ioc_enclave_provision
102
103Enclave runtime management
104--------------------------
105
106Systems supporting SGX2 additionally support changes to initialized
107enclaves: modifying enclave page permissions and type, and dynamically
108adding and removing of enclave pages. When an enclave accesses an address
109within its address range that does not have a backing page then a new
110regular page will be dynamically added to the enclave. The enclave is
111still required to run EACCEPT on the new page before it can be used.
112
113.. kernel-doc:: arch/x86/kernel/cpu/sgx/ioctl.c
114   :functions: sgx_ioc_enclave_restrict_permissions
115               sgx_ioc_enclave_modify_types
116               sgx_ioc_enclave_remove_pages
117
118Enclave vDSO
119------------
120
121Entering an enclave can only be done through SGX-specific EENTER and ERESUME
122functions, and is a non-trivial process.  Because of the complexity of
123transitioning to and from an enclave, enclaves typically utilize a library to
124handle the actual transitions.  This is roughly analogous to how glibc
125implementations are used by most applications to wrap system calls.
126
127Another crucial characteristic of enclaves is that they can generate exceptions
128as part of their normal operation that need to be handled in the enclave or are
129unique to SGX.
130
131Instead of the traditional signal mechanism to handle these exceptions, SGX
132can leverage special exception fixup provided by the vDSO.  The kernel-provided
133vDSO function wraps low-level transitions to/from the enclave like EENTER and
134ERESUME.  The vDSO function intercepts exceptions that would otherwise generate
135a signal and return the fault information directly to its caller.  This avoids
136the need to juggle signal handlers.
137
138.. kernel-doc:: arch/x86/include/uapi/asm/sgx.h
139   :functions: vdso_sgx_enter_enclave_t
140
141ksgxd
142=====
143
144SGX support includes a kernel thread called *ksgxd*.
145
146EPC sanitization
147----------------
148
149ksgxd is started when SGX initializes.  Enclave memory is typically ready
150for use when the processor powers on or resets.  However, if SGX has been in
151use since the reset, enclave pages may be in an inconsistent state.  This might
152occur after a crash and kexec() cycle, for instance.  At boot, ksgxd
153reinitializes all enclave pages so that they can be allocated and re-used.
154
155The sanitization is done by going through EPC address space and applying the
156EREMOVE function to each physical page. Some enclave pages like SECS pages have
157hardware dependencies on other pages which prevents EREMOVE from functioning.
158Executing two EREMOVE passes removes the dependencies.
159
160Page reclaimer
161--------------
162
163Similar to the core kswapd, ksgxd, is responsible for managing the
164overcommitment of enclave memory.  If the system runs out of enclave memory,
165*ksgxd* “swaps” enclave memory to normal memory.
166
167Launch Control
168==============
169
170SGX provides a launch control mechanism. After all enclave pages have been
171copied, kernel executes EINIT function, which initializes the enclave. Only after
172this the CPU can execute inside the enclave.
173
174EINIT function takes an RSA-3072 signature of the enclave measurement.  The function
175checks that the measurement is correct and signature is signed with the key
176hashed to the four **IA32_SGXLEPUBKEYHASH{0, 1, 2, 3}** MSRs representing the
177SHA256 of a public key.
178
179Those MSRs can be configured by the BIOS to be either readable or writable.
180Linux supports only writable configuration in order to give full control to the
181kernel on launch control policy. Before calling EINIT function, the driver sets
182the MSRs to match the enclave's signing key.
183
184Encryption engines
185==================
186
187In order to conceal the enclave data while it is out of the CPU package, the
188memory controller has an encryption engine to transparently encrypt and decrypt
189enclave memory.
190
191In CPUs prior to Ice Lake, the Memory Encryption Engine (MEE) is used to
192encrypt pages leaving the CPU caches. MEE uses a n-ary Merkle tree with root in
193SRAM to maintain integrity of the encrypted data. This provides integrity and
194anti-replay protection but does not scale to large memory sizes because the time
195required to update the Merkle tree grows logarithmically in relation to the
196memory size.
197
198CPUs starting from Icelake use Total Memory Encryption (TME) in the place of
199MEE. TME-based SGX implementations do not have an integrity Merkle tree, which
200means integrity and replay-attacks are not mitigated.  B, it includes
201additional changes to prevent cipher text from being returned and SW memory
202aliases from being created.
203
204DMA to enclave memory is blocked by range registers on both MEE and TME systems
205(SDM section 41.10).
206
207Usage Models
208============
209
210Shared Library
211--------------
212
213Sensitive data and the code that acts on it is partitioned from the application
214into a separate library. The library is then linked as a DSO which can be loaded
215into an enclave. The application can then make individual function calls into
216the enclave through special SGX instructions. A run-time within the enclave is
217configured to marshal function parameters into and out of the enclave and to
218call the correct library function.
219
220Application Container
221---------------------
222
223An application may be loaded into a container enclave which is specially
224configured with a library OS and run-time which permits the application to run.
225The enclave run-time and library OS work together to execute the application
226when a thread enters the enclave.
227
228Impact of Potential Kernel SGX Bugs
229===================================
230
231EPC leaks
232---------
233
234When EPC page leaks happen, a WARNING like this is shown in dmesg:
235
236"EREMOVE returned ... and an EPC page was leaked.  SGX may become unusable..."
237
238This is effectively a kernel use-after-free of an EPC page, and due
239to the way SGX works, the bug is detected at freeing. Rather than
240adding the page back to the pool of available EPC pages, the kernel
241intentionally leaks the page to avoid additional errors in the future.
242
243When this happens, the kernel will likely soon leak more EPC pages, and
244SGX will likely become unusable because the memory available to SGX is
245limited. However, while this may be fatal to SGX, the rest of the kernel
246is unlikely to be impacted and should continue to work.
247
248As a result, when this happpens, user should stop running any new
249SGX workloads, (or just any new workloads), and migrate all valuable
250workloads. Although a machine reboot can recover all EPC memory, the bug
251should be reported to Linux developers.
252
253
254Virtual EPC
255===========
256
257The implementation has also a virtual EPC driver to support SGX enclaves
258in guests. Unlike the SGX driver, an EPC page allocated by the virtual
259EPC driver doesn't have a specific enclave associated with it. This is
260because KVM doesn't track how a guest uses EPC pages.
261
262As a result, the SGX core page reclaimer doesn't support reclaiming EPC
263pages allocated to KVM guests through the virtual EPC driver. If the
264user wants to deploy SGX applications both on the host and in guests
265on the same machine, the user should reserve enough EPC (by taking out
266total virtual EPC size of all SGX VMs from the physical EPC size) for
267host SGX applications so they can run with acceptable performance.
268
269Architectural behavior is to restore all EPC pages to an uninitialized
270state also after a guest reboot.  Because this state can be reached only
271through the privileged ``ENCLS[EREMOVE]`` instruction, ``/dev/sgx_vepc``
272provides the ``SGX_IOC_VEPC_REMOVE_ALL`` ioctl to execute the instruction
273on all pages in the virtual EPC.
274
275``EREMOVE`` can fail for three reasons.  Userspace must pay attention
276to expected failures and handle them as follows:
277
2781. Page removal will always fail when any thread is running in the
279   enclave to which the page belongs.  In this case the ioctl will
280   return ``EBUSY`` independent of whether it has successfully removed
281   some pages; userspace can avoid these failures by preventing execution
282   of any vcpu which maps the virtual EPC.
283
2842. Page removal will cause a general protection fault if two calls to
285   ``EREMOVE`` happen concurrently for pages that refer to the same
286   "SECS" metadata pages.  This can happen if there are concurrent
287   invocations to ``SGX_IOC_VEPC_REMOVE_ALL``, or if a ``/dev/sgx_vepc``
288   file descriptor in the guest is closed at the same time as
289   ``SGX_IOC_VEPC_REMOVE_ALL``; it will also be reported as ``EBUSY``.
290   This can be avoided in userspace by serializing calls to the ioctl()
291   and to close(), but in general it should not be a problem.
292
2933. Finally, page removal will fail for SECS metadata pages which still
294   have child pages.  Child pages can be removed by executing
295   ``SGX_IOC_VEPC_REMOVE_ALL`` on all ``/dev/sgx_vepc`` file descriptors
296   mapped into the guest.  This means that the ioctl() must be called
297   twice: an initial set of calls to remove child pages and a subsequent
298   set of calls to remove SECS pages.  The second set of calls is only
299   required for those mappings that returned a nonzero value from the
300   first call.  It indicates a bug in the kernel or the userspace client
301   if any of the second round of ``SGX_IOC_VEPC_REMOVE_ALL`` calls has
302   a return code other than 0.
303