1 # SPDX-License-Identifier: GPL-2.0-only
2 menu "Kernel hardening options"
4 config GCC_PLUGIN_STRUCTLEAK
7 While the kernel is built with warnings enabled for any missed
8 stack variable initializations, this warning is silenced for
9 anything passed by reference to another function, under the
10 occasionally misguided assumption that the function will do
11 the initialization. As this regularly leads to exploitable
12 flaws, this plugin is available to identify and zero-initialize
13 such variables, depending on the chosen level of coverage.
15 This plugin was originally ported from grsecurity/PaX. More
17 * https://grsecurity.net/
18 * https://pax.grsecurity.net/
20 menu "Memory initialization"
22 config CC_HAS_AUTO_VAR_INIT_PATTERN
23 def_bool $(cc-option,-ftrivial-auto-var-init=pattern)
25 config CC_HAS_AUTO_VAR_INIT_ZERO
26 # GCC ignores the -enable flag, so we can test for the feature with
27 # a single invocation using the flag, but drop it as appropriate in
28 # the Makefile, depending on the presence of Clang.
29 def_bool $(cc-option,-ftrivial-auto-var-init=zero -enable-trivial-auto-var-init-zero-knowing-it-will-be-removed-from-clang)
32 prompt "Initialize kernel stack variables at function entry"
33 default GCC_PLUGIN_STRUCTLEAK_BYREF_ALL if COMPILE_TEST && GCC_PLUGINS
34 default INIT_STACK_ALL_PATTERN if COMPILE_TEST && CC_HAS_AUTO_VAR_INIT_PATTERN
35 default INIT_STACK_ALL_ZERO if CC_HAS_AUTO_VAR_INIT_ZERO
36 default INIT_STACK_NONE
38 This option enables initialization of stack variables at
39 function entry time. This has the possibility to have the
40 greatest coverage (since all functions can have their
41 variables initialized), but the performance impact depends
42 on the function calling complexity of a given workload's
45 This chooses the level of coverage over classes of potentially
46 uninitialized variables. The selected class of variable will be
47 initialized before use in a function.
49 config INIT_STACK_NONE
50 bool "no automatic stack variable initialization (weakest)"
52 Disable automatic stack variable initialization.
53 This leaves the kernel vulnerable to the standard
54 classes of uninitialized stack variable exploits
55 and information exposures.
57 config GCC_PLUGIN_STRUCTLEAK_USER
58 bool "zero-init structs marked for userspace (weak)"
59 # Plugin can be removed once the kernel only supports GCC 12+
60 depends on GCC_PLUGINS && !CC_HAS_AUTO_VAR_INIT_ZERO
61 select GCC_PLUGIN_STRUCTLEAK
63 Zero-initialize any structures on the stack containing
64 a __user attribute. This can prevent some classes of
65 uninitialized stack variable exploits and information
66 exposures, like CVE-2013-2141:
67 https://git.kernel.org/linus/b9e146d8eb3b9eca
69 config GCC_PLUGIN_STRUCTLEAK_BYREF
70 bool "zero-init structs passed by reference (strong)"
71 # Plugin can be removed once the kernel only supports GCC 12+
72 depends on GCC_PLUGINS && !CC_HAS_AUTO_VAR_INIT_ZERO
73 depends on !(KASAN && KASAN_STACK)
74 select GCC_PLUGIN_STRUCTLEAK
76 Zero-initialize any structures on the stack that may
77 be passed by reference and had not already been
78 explicitly initialized. This can prevent most classes
79 of uninitialized stack variable exploits and information
80 exposures, like CVE-2017-1000410:
81 https://git.kernel.org/linus/06e7e776ca4d3654
83 As a side-effect, this keeps a lot of variables on the
84 stack that can otherwise be optimized out, so combining
85 this with CONFIG_KASAN_STACK can lead to a stack overflow
88 config GCC_PLUGIN_STRUCTLEAK_BYREF_ALL
89 bool "zero-init everything passed by reference (very strong)"
90 # Plugin can be removed once the kernel only supports GCC 12+
91 depends on GCC_PLUGINS && !CC_HAS_AUTO_VAR_INIT_ZERO
92 depends on !(KASAN && KASAN_STACK)
93 select GCC_PLUGIN_STRUCTLEAK
95 Zero-initialize any stack variables that may be passed
96 by reference and had not already been explicitly
97 initialized. This is intended to eliminate all classes
98 of uninitialized stack variable exploits and information
101 As a side-effect, this keeps a lot of variables on the
102 stack that can otherwise be optimized out, so combining
103 this with CONFIG_KASAN_STACK can lead to a stack overflow
106 config INIT_STACK_ALL_PATTERN
107 bool "pattern-init everything (strongest)"
108 depends on CC_HAS_AUTO_VAR_INIT_PATTERN
111 Initializes everything on the stack (including padding)
112 with a specific debug value. This is intended to eliminate
113 all classes of uninitialized stack variable exploits and
114 information exposures, even variables that were warned about
115 having been left uninitialized.
117 Pattern initialization is known to provoke many existing bugs
118 related to uninitialized locals, e.g. pointers receive
119 non-NULL values, buffer sizes and indices are very big. The
120 pattern is situation-specific; Clang on 64-bit uses 0xAA
121 repeating for all types and padding except float and double
122 which use 0xFF repeating (-NaN). Clang on 32-bit uses 0xFF
123 repeating for all types and padding.
125 config INIT_STACK_ALL_ZERO
126 bool "zero-init everything (strongest and safest)"
127 depends on CC_HAS_AUTO_VAR_INIT_ZERO
130 Initializes everything on the stack (including padding)
131 with a zero value. This is intended to eliminate all
132 classes of uninitialized stack variable exploits and
133 information exposures, even variables that were warned
134 about having been left uninitialized.
136 Zero initialization provides safe defaults for strings
137 (immediately NUL-terminated), pointers (NULL), indices
138 (index 0), and sizes (0 length), so it is therefore more
139 suitable as a production security mitigation than pattern
144 config GCC_PLUGIN_STRUCTLEAK_VERBOSE
145 bool "Report forcefully initialized variables"
146 depends on GCC_PLUGIN_STRUCTLEAK
147 depends on !COMPILE_TEST # too noisy
149 This option will cause a warning to be printed each time the
150 structleak plugin finds a variable it thinks needs to be
151 initialized. Since not all existing initializers are detected
152 by the plugin, this can produce false positive warnings.
154 config GCC_PLUGIN_STACKLEAK
155 bool "Poison kernel stack before returning from syscalls"
156 depends on GCC_PLUGINS
157 depends on HAVE_ARCH_STACKLEAK
159 This option makes the kernel erase the kernel stack before
160 returning from system calls. This has the effect of leaving
161 the stack initialized to the poison value, which both reduces
162 the lifetime of any sensitive stack contents and reduces
163 potential for uninitialized stack variable exploits or information
164 exposures (it does not cover functions reaching the same stack
165 depth as prior functions during the same syscall). This blocks
166 most uninitialized stack variable attacks, with the performance
167 impact being driven by the depth of the stack usage, rather than
168 the function calling complexity.
170 The performance impact on a single CPU system kernel compilation
171 sees a 1% slowdown, other systems and workloads may vary and you
172 are advised to test this feature on your expected workload before
175 This plugin was ported from grsecurity/PaX. More information at:
176 * https://grsecurity.net/
177 * https://pax.grsecurity.net/
179 config GCC_PLUGIN_STACKLEAK_VERBOSE
180 bool "Report stack depth analysis instrumentation" if EXPERT
181 depends on GCC_PLUGIN_STACKLEAK
182 depends on !COMPILE_TEST # too noisy
184 This option will cause a warning to be printed each time the
185 stackleak plugin finds a function it thinks needs to be
186 instrumented. This is useful for comparing coverage between
189 config STACKLEAK_TRACK_MIN_SIZE
190 int "Minimum stack frame size of functions tracked by STACKLEAK"
193 depends on GCC_PLUGIN_STACKLEAK
195 The STACKLEAK gcc plugin instruments the kernel code for tracking
196 the lowest border of the kernel stack (and for some other purposes).
197 It inserts the stackleak_track_stack() call for the functions with
198 a stack frame size greater than or equal to this parameter.
199 If unsure, leave the default value 100.
201 config STACKLEAK_METRICS
202 bool "Show STACKLEAK metrics in the /proc file system"
203 depends on GCC_PLUGIN_STACKLEAK
206 If this is set, STACKLEAK metrics for every task are available in
207 the /proc file system. In particular, /proc/<pid>/stack_depth
208 shows the maximum kernel stack consumption for the current and
209 previous syscalls. Although this information is not precise, it
210 can be useful for estimating the STACKLEAK performance impact for
213 config STACKLEAK_RUNTIME_DISABLE
214 bool "Allow runtime disabling of kernel stack erasing"
215 depends on GCC_PLUGIN_STACKLEAK
217 This option provides 'stack_erasing' sysctl, which can be used in
218 runtime to control kernel stack erasing for kernels built with
219 CONFIG_GCC_PLUGIN_STACKLEAK.
221 config INIT_ON_ALLOC_DEFAULT_ON
222 bool "Enable heap memory zeroing on allocation by default"
225 This has the effect of setting "init_on_alloc=1" on the kernel
226 command line. This can be disabled with "init_on_alloc=0".
227 When "init_on_alloc" is enabled, all page allocator and slab
228 allocator memory will be zeroed when allocated, eliminating
229 many kinds of "uninitialized heap memory" flaws, especially
230 heap content exposures. The performance impact varies by
231 workload, but most cases see <1% impact. Some synthetic
232 workloads have measured as high as 7%.
234 config INIT_ON_FREE_DEFAULT_ON
235 bool "Enable heap memory zeroing on free by default"
238 This has the effect of setting "init_on_free=1" on the kernel
239 command line. This can be disabled with "init_on_free=0".
240 Similar to "init_on_alloc", when "init_on_free" is enabled,
241 all page allocator and slab allocator memory will be zeroed
242 when freed, eliminating many kinds of "uninitialized heap memory"
243 flaws, especially heap content exposures. The primary difference
244 with "init_on_free" is that data lifetime in memory is reduced,
245 as anything freed is wiped immediately, making live forensics or
246 cold boot memory attacks unable to recover freed memory contents.
247 The performance impact varies by workload, but is more expensive
248 than "init_on_alloc" due to the negative cache effects of
249 touching "cold" memory areas. Most cases see 3-5% impact. Some
250 synthetic workloads have measured as high as 8%.
252 config CC_HAS_ZERO_CALL_USED_REGS
253 def_bool $(cc-option,-fzero-call-used-regs=used-gpr)
255 config ZERO_CALL_USED_REGS
256 bool "Enable register zeroing on function exit"
257 depends on CC_HAS_ZERO_CALL_USED_REGS
259 At the end of functions, always zero any caller-used register
260 contents. This helps ensure that temporary values are not
261 leaked beyond the function boundary. This means that register
262 contents are less likely to be available for side channels
263 and information exposures. Additionally, this helps reduce the
264 number of useful ROP gadgets by about 20% (and removes compiler
265 generated "write-what-where" gadgets) in the resulting kernel
266 image. This has a less than 1% performance impact on most
267 workloads. Image size growth depends on architecture, and should
268 be evaluated for suitability. For example, x86_64 grows by less
269 than 1%, and arm64 grows by about 5%.
273 config CC_HAS_RANDSTRUCT
274 def_bool $(cc-option,-frandomize-layout-seed-file=/dev/null)
277 prompt "Randomize layout of sensitive kernel structures"
278 default RANDSTRUCT_FULL if COMPILE_TEST && (GCC_PLUGINS || CC_HAS_RANDSTRUCT)
279 default RANDSTRUCT_NONE
281 If you enable this, the layouts of structures that are entirely
282 function pointers (and have not been manually annotated with
283 __no_randomize_layout), or structures that have been explicitly
284 marked with __randomize_layout, will be randomized at compile-time.
285 This can introduce the requirement of an additional information
286 exposure vulnerability for exploits targeting these structure
289 Enabling this feature will introduce some performance impact,
290 slightly increase memory usage, and prevent the use of forensic
291 tools like Volatility against the system (unless the kernel
292 source tree isn't cleaned after kernel installation).
294 The seed used for compilation is in scripts/basic/randomize.seed.
295 It remains after a "make clean" to allow for external modules to
296 be compiled with the existing seed and will be removed by a
297 "make mrproper" or "make distclean". This file should not be made
298 public, or the structure layout can be determined.
300 config RANDSTRUCT_NONE
301 bool "Disable structure layout randomization"
303 Build normally: no structure layout randomization.
305 config RANDSTRUCT_FULL
306 bool "Fully randomize structure layout"
307 depends on CC_HAS_RANDSTRUCT || GCC_PLUGINS
308 select MODVERSIONS if MODULES
310 Fully randomize the member layout of sensitive
311 structures as much as possible, which may have both a
312 memory size and performance impact.
314 One difference between the Clang and GCC plugin
315 implementations is the handling of bitfields. The GCC
316 plugin treats them as fully separate variables,
317 introducing sometimes significant padding. Clang tries
318 to keep adjacent bitfields together, but with their bit
321 config RANDSTRUCT_PERFORMANCE
322 bool "Limit randomization of structure layout to cache-lines"
323 depends on GCC_PLUGINS
324 select MODVERSIONS if MODULES
326 Randomization of sensitive kernel structures will make a
327 best effort at restricting randomization to cacheline-sized
328 groups of members. It will further not randomize bitfields
329 in structures. This reduces the performance hit of RANDSTRUCT
330 at the cost of weakened randomization.
334 def_bool !RANDSTRUCT_NONE
336 config GCC_PLUGIN_RANDSTRUCT
337 def_bool GCC_PLUGINS && RANDSTRUCT
339 Use GCC plugin to randomize structure layout.
341 This plugin was ported from grsecurity/PaX. More
343 * https://grsecurity.net/
344 * https://pax.grsecurity.net/