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 depends on GCC_PLUGINS
60 select GCC_PLUGIN_STRUCTLEAK
62 Zero-initialize any structures on the stack containing
63 a __user attribute. This can prevent some classes of
64 uninitialized stack variable exploits and information
65 exposures, like CVE-2013-2141:
66 https://git.kernel.org/linus/b9e146d8eb3b9eca
68 config GCC_PLUGIN_STRUCTLEAK_BYREF
69 bool "zero-init structs passed by reference (strong)"
70 depends on GCC_PLUGINS
71 depends on !(KASAN && KASAN_STACK)
72 select GCC_PLUGIN_STRUCTLEAK
74 Zero-initialize any structures on the stack that may
75 be passed by reference and had not already been
76 explicitly initialized. This can prevent most classes
77 of uninitialized stack variable exploits and information
78 exposures, like CVE-2017-1000410:
79 https://git.kernel.org/linus/06e7e776ca4d3654
81 As a side-effect, this keeps a lot of variables on the
82 stack that can otherwise be optimized out, so combining
83 this with CONFIG_KASAN_STACK can lead to a stack overflow
86 config GCC_PLUGIN_STRUCTLEAK_BYREF_ALL
87 bool "zero-init everything passed by reference (very strong)"
88 depends on GCC_PLUGINS
89 depends on !(KASAN && KASAN_STACK)
90 select GCC_PLUGIN_STRUCTLEAK
92 Zero-initialize any stack variables that may be passed
93 by reference and had not already been explicitly
94 initialized. This is intended to eliminate all classes
95 of uninitialized stack variable exploits and information
98 As a side-effect, this keeps a lot of variables on the
99 stack that can otherwise be optimized out, so combining
100 this with CONFIG_KASAN_STACK can lead to a stack overflow
103 config INIT_STACK_ALL_PATTERN
104 bool "pattern-init everything (strongest)"
105 depends on CC_HAS_AUTO_VAR_INIT_PATTERN
107 Initializes everything on the stack (including padding)
108 with a specific debug value. This is intended to eliminate
109 all classes of uninitialized stack variable exploits and
110 information exposures, even variables that were warned about
111 having been left uninitialized.
113 Pattern initialization is known to provoke many existing bugs
114 related to uninitialized locals, e.g. pointers receive
115 non-NULL values, buffer sizes and indices are very big. The
116 pattern is situation-specific; Clang on 64-bit uses 0xAA
117 repeating for all types and padding except float and double
118 which use 0xFF repeating (-NaN). Clang on 32-bit uses 0xFF
119 repeating for all types and padding.
121 config INIT_STACK_ALL_ZERO
122 bool "zero-init everything (strongest and safest)"
123 depends on CC_HAS_AUTO_VAR_INIT_ZERO
125 Initializes everything on the stack (including padding)
126 with a zero value. This is intended to eliminate all
127 classes of uninitialized stack variable exploits and
128 information exposures, even variables that were warned
129 about having been left uninitialized.
131 Zero initialization provides safe defaults for strings
132 (immediately NUL-terminated), pointers (NULL), indices
133 (index 0), and sizes (0 length), so it is therefore more
134 suitable as a production security mitigation than pattern
139 config GCC_PLUGIN_STRUCTLEAK_VERBOSE
140 bool "Report forcefully initialized variables"
141 depends on GCC_PLUGIN_STRUCTLEAK
142 depends on !COMPILE_TEST # too noisy
144 This option will cause a warning to be printed each time the
145 structleak plugin finds a variable it thinks needs to be
146 initialized. Since not all existing initializers are detected
147 by the plugin, this can produce false positive warnings.
149 config GCC_PLUGIN_STACKLEAK
150 bool "Poison kernel stack before returning from syscalls"
151 depends on GCC_PLUGINS
152 depends on HAVE_ARCH_STACKLEAK
154 This option makes the kernel erase the kernel stack before
155 returning from system calls. This has the effect of leaving
156 the stack initialized to the poison value, which both reduces
157 the lifetime of any sensitive stack contents and reduces
158 potential for uninitialized stack variable exploits or information
159 exposures (it does not cover functions reaching the same stack
160 depth as prior functions during the same syscall). This blocks
161 most uninitialized stack variable attacks, with the performance
162 impact being driven by the depth of the stack usage, rather than
163 the function calling complexity.
165 The performance impact on a single CPU system kernel compilation
166 sees a 1% slowdown, other systems and workloads may vary and you
167 are advised to test this feature on your expected workload before
170 This plugin was ported from grsecurity/PaX. More information at:
171 * https://grsecurity.net/
172 * https://pax.grsecurity.net/
174 config STACKLEAK_TRACK_MIN_SIZE
175 int "Minimum stack frame size of functions tracked by STACKLEAK"
178 depends on GCC_PLUGIN_STACKLEAK
180 The STACKLEAK gcc plugin instruments the kernel code for tracking
181 the lowest border of the kernel stack (and for some other purposes).
182 It inserts the stackleak_track_stack() call for the functions with
183 a stack frame size greater than or equal to this parameter.
184 If unsure, leave the default value 100.
186 config STACKLEAK_METRICS
187 bool "Show STACKLEAK metrics in the /proc file system"
188 depends on GCC_PLUGIN_STACKLEAK
191 If this is set, STACKLEAK metrics for every task are available in
192 the /proc file system. In particular, /proc/<pid>/stack_depth
193 shows the maximum kernel stack consumption for the current and
194 previous syscalls. Although this information is not precise, it
195 can be useful for estimating the STACKLEAK performance impact for
198 config STACKLEAK_RUNTIME_DISABLE
199 bool "Allow runtime disabling of kernel stack erasing"
200 depends on GCC_PLUGIN_STACKLEAK
202 This option provides 'stack_erasing' sysctl, which can be used in
203 runtime to control kernel stack erasing for kernels built with
204 CONFIG_GCC_PLUGIN_STACKLEAK.
206 config INIT_ON_ALLOC_DEFAULT_ON
207 bool "Enable heap memory zeroing on allocation by default"
209 This has the effect of setting "init_on_alloc=1" on the kernel
210 command line. This can be disabled with "init_on_alloc=0".
211 When "init_on_alloc" is enabled, all page allocator and slab
212 allocator memory will be zeroed when allocated, eliminating
213 many kinds of "uninitialized heap memory" flaws, especially
214 heap content exposures. The performance impact varies by
215 workload, but most cases see <1% impact. Some synthetic
216 workloads have measured as high as 7%.
218 config INIT_ON_FREE_DEFAULT_ON
219 bool "Enable heap memory zeroing on free by default"
221 This has the effect of setting "init_on_free=1" on the kernel
222 command line. This can be disabled with "init_on_free=0".
223 Similar to "init_on_alloc", when "init_on_free" is enabled,
224 all page allocator and slab allocator memory will be zeroed
225 when freed, eliminating many kinds of "uninitialized heap memory"
226 flaws, especially heap content exposures. The primary difference
227 with "init_on_free" is that data lifetime in memory is reduced,
228 as anything freed is wiped immediately, making live forensics or
229 cold boot memory attacks unable to recover freed memory contents.
230 The performance impact varies by workload, but is more expensive
231 than "init_on_alloc" due to the negative cache effects of
232 touching "cold" memory areas. Most cases see 3-5% impact. Some
233 synthetic workloads have measured as high as 8%.
235 config CC_HAS_ZERO_CALL_USED_REGS
236 def_bool $(cc-option,-fzero-call-used-regs=used-gpr)
238 config ZERO_CALL_USED_REGS
239 bool "Enable register zeroing on function exit"
240 depends on CC_HAS_ZERO_CALL_USED_REGS
242 At the end of functions, always zero any caller-used register
243 contents. This helps ensure that temporary values are not
244 leaked beyond the function boundary. This means that register
245 contents are less likely to be available for side channels
246 and information exposures. Additionally, this helps reduce the
247 number of useful ROP gadgets by about 20% (and removes compiler
248 generated "write-what-where" gadgets) in the resulting kernel
249 image. This has a less than 1% performance impact on most
250 workloads. Image size growth depends on architecture, and should
251 be evaluated for suitability. For example, x86_64 grows by less
252 than 1%, and arm64 grows by about 5%.