1 # SPDX-License-Identifier: GPL-2.0-only
2
3 config HAVE_ARCH_KCSAN
4 bool
5
6 config HAVE_KCSAN_COMPILER
7 def_bool (CC_IS_CLANG && $(cc-option,-fsanitize=thread -mllvm -tsan-distinguish-volatile=1)) || \
8 (CC_IS_GCC && $(cc-option,-fsanitize=thread --param tsan-distinguish-volatile=1))
9 help
10 For the list of compilers that support KCSAN, please see
11 <file:Documentation/dev-tools/kcsan.rst>.
12
13 menuconfig KCSAN
14 bool "KCSAN: dynamic data race detector"
15 depends on HAVE_ARCH_KCSAN && HAVE_KCSAN_COMPILER
16 depends on DEBUG_KERNEL && !KASAN
17 select CONSTRUCTORS
18 select STACKTRACE
19 help
20 The Kernel Concurrency Sanitizer (KCSAN) is a dynamic
21 data-race detector that relies on compile-time instrumentation.
22 KCSAN uses a watchpoint-based sampling approach to detect races.
23
24 While KCSAN's primary purpose is to detect data races, it
25 also provides assertions to check data access constraints.
26 These assertions can expose bugs that do not manifest as
27 data races.
28
29 See <file:Documentation/dev-tools/kcsan.rst> for more details.
30
31 if KCSAN
32
33 config CC_HAS_TSAN_COMPOUND_READ_BEFORE_WRITE
34 def_bool (CC_IS_CLANG && $(cc-option,-fsanitize=thread -mllvm -tsan-compound-read-before-write=1)) || \
35 (CC_IS_GCC && $(cc-option,-fsanitize=thread --param tsan-compound-read-before-write=1))
36 help
37 The compiler instruments plain compound read-write operations
38 differently (++, --, +=, -=, |=, &=, etc.), which allows KCSAN to
39 distinguish them from other plain accesses. This is currently
40 supported by Clang 12 or later.
41
42 config KCSAN_VERBOSE
43 bool "Show verbose reports with more information about system state"
44 depends on PROVE_LOCKING
45 help
46 If enabled, reports show more information about the system state that
47 may help better analyze and debug races. This includes held locks and
48 IRQ trace events.
49
50 While this option should generally be benign, we call into more
51 external functions on report generation; if a race report is
52 generated from any one of them, system stability may suffer due to
53 deadlocks or recursion. If in doubt, say N.
54
55 config KCSAN_SELFTEST
56 bool "Perform short selftests on boot"
57 default y
58 help
59 Run KCSAN selftests on boot. On test failure, causes the kernel to
60 panic. Recommended to be enabled, ensuring critical functionality
61 works as intended.
62
63 config KCSAN_KUNIT_TEST
64 tristate "KCSAN test for integrated runtime behaviour" if !KUNIT_ALL_TESTS
65 default KUNIT_ALL_TESTS
66 depends on TRACEPOINTS && KUNIT
67 select TORTURE_TEST
68 help
69 KCSAN test focusing on behaviour of the integrated runtime. Tests
70 various race scenarios, and verifies the reports generated to
71 console. Makes use of KUnit for test organization, and the Torture
72 framework for test thread control.
73
74 Each test case may run at least up to KCSAN_REPORT_ONCE_IN_MS
75 milliseconds. Test run duration may be optimized by building the
76 kernel and KCSAN test with KCSAN_REPORT_ONCE_IN_MS set to a lower
77 than default value.
78
79 Say Y here if you want the test to be built into the kernel and run
80 during boot; say M if you want the test to build as a module; say N
81 if you are unsure.
82
83 config KCSAN_EARLY_ENABLE
84 bool "Early enable during boot"
85 default y
86 help
87 If KCSAN should be enabled globally as soon as possible. KCSAN can
88 later be enabled/disabled via debugfs.
89
90 config KCSAN_NUM_WATCHPOINTS
91 int "Number of available watchpoints"
92 default 64
93 help
94 Total number of available watchpoints. An address range maps into a
95 specific watchpoint slot as specified in kernel/kcsan/encoding.h.
96 Although larger number of watchpoints may not be usable due to
97 limited number of CPUs, a larger value helps to improve performance
98 due to reducing cache-line contention. The chosen default is a
99 conservative value; we should almost never observe "no_capacity"
100 events (see /sys/kernel/debug/kcsan).
101
102 config KCSAN_UDELAY_TASK
103 int "Delay in microseconds (for tasks)"
104 default 80
105 help
106 For tasks, the microsecond delay after setting up a watchpoint.
107
108 config KCSAN_UDELAY_INTERRUPT
109 int "Delay in microseconds (for interrupts)"
110 default 20
111 help
112 For interrupts, the microsecond delay after setting up a watchpoint.
113 Interrupts have tighter latency requirements, and their delay should
114 be lower than for tasks.
115
116 config KCSAN_DELAY_RANDOMIZE
117 bool "Randomize above delays"
118 default y
119 help
120 If delays should be randomized, where the maximum is KCSAN_UDELAY_*.
121 If false, the chosen delays are always the KCSAN_UDELAY_* values
122 as defined above.
123
124 config KCSAN_SKIP_WATCH
125 int "Skip instructions before setting up watchpoint"
126 default 4000
127 help
128 The number of per-CPU memory operations to skip, before another
129 watchpoint is set up, i.e. one in KCSAN_SKIP_WATCH per-CPU
130 memory operations are used to set up a watchpoint. A smaller value
131 results in more aggressive race detection, whereas a larger value
132 improves system performance at the cost of missing some races.
133
134 config KCSAN_SKIP_WATCH_RANDOMIZE
135 bool "Randomize watchpoint instruction skip count"
136 default y
137 help
138 If instruction skip count should be randomized, where the maximum is
139 KCSAN_SKIP_WATCH. If false, the chosen value is always
140 KCSAN_SKIP_WATCH.
141
142 config KCSAN_INTERRUPT_WATCHER
143 bool "Interruptible watchers" if !KCSAN_STRICT
144 default KCSAN_STRICT
145 help
146 If enabled, a task that set up a watchpoint may be interrupted while
147 delayed. This option will allow KCSAN to detect races between
148 interrupted tasks and other threads of execution on the same CPU.
149
150 Currently disabled by default, because not all safe per-CPU access
151 primitives and patterns may be accounted for, and therefore could
152 result in false positives.
153
154 config KCSAN_REPORT_ONCE_IN_MS
155 int "Duration in milliseconds, in which any given race is only reported once"
156 default 3000
157 help
158 Any given race is only reported once in the defined time window.
159 Different races may still generate reports within a duration that is
160 smaller than the duration defined here. This allows rate limiting
161 reporting to avoid flooding the console with reports. Setting this
162 to 0 disables rate limiting.
163
164 # The main purpose of the below options is to control reported data races, and
165 # are not expected to be switched frequently by non-testers or at runtime.
166 # The defaults are chosen to be conservative, and can miss certain bugs.
167
168 config KCSAN_REPORT_RACE_UNKNOWN_ORIGIN
169 bool "Report races of unknown origin"
170 default y
171 help
172 If KCSAN should report races where only one access is known, and the
173 conflicting access is of unknown origin. This type of race is
174 reported if it was only possible to infer a race due to a data value
175 change while an access is being delayed on a watchpoint.
176
177 config KCSAN_STRICT
178 bool "Strict data-race checking"
179 help
180 KCSAN will report data races with the strictest possible rules, which
181 closely aligns with the rules defined by the Linux-kernel memory
182 consistency model (LKMM).
183
184 config KCSAN_WEAK_MEMORY
185 bool "Enable weak memory modeling to detect missing memory barriers"
186 default y
187 depends on KCSAN_STRICT
188 # We can either let objtool nop __tsan_func_{entry,exit}() and builtin
189 # atomics instrumentation in .noinstr.text, or use a compiler that can
190 # implement __no_kcsan to really remove all instrumentation.
191 depends on !ARCH_WANTS_NO_INSTR || HAVE_NOINSTR_HACK || \
192 CC_IS_GCC || CLANG_VERSION >= 140000
193 select OBJTOOL if HAVE_NOINSTR_HACK
194 help
195 Enable support for modeling a subset of weak memory, which allows
196 detecting a subset of data races due to missing memory barriers.
197
198 Depends on KCSAN_STRICT, because the options strengthening certain
199 plain accesses by default (depending on !KCSAN_STRICT) reduce the
200 ability to detect any data races invoving reordered accesses, in
201 particular reordered writes.
202
203 Weak memory modeling relies on additional instrumentation and may
204 affect performance.
205
206 config KCSAN_REPORT_VALUE_CHANGE_ONLY
207 bool "Only report races where watcher observed a data value change"
208 default y
209 depends on !KCSAN_STRICT
210 help
211 If enabled and a conflicting write is observed via a watchpoint, but
212 the data value of the memory location was observed to remain
213 unchanged, do not report the data race.
214
215 config KCSAN_ASSUME_PLAIN_WRITES_ATOMIC
216 bool "Assume that plain aligned writes up to word size are atomic"
217 default y
218 depends on !KCSAN_STRICT
219 help
220 Assume that plain aligned writes up to word size are atomic by
221 default, and also not subject to other unsafe compiler optimizations
222 resulting in data races. This will cause KCSAN to not report data
223 races due to conflicts where the only plain accesses are aligned
224 writes up to word size: conflicts between marked reads and plain
225 aligned writes up to word size will not be reported as data races;
226 notice that data races between two conflicting plain aligned writes
227 will also not be reported.
228
229 config KCSAN_IGNORE_ATOMICS
230 bool "Do not instrument marked atomic accesses"
231 depends on !KCSAN_STRICT
232 help
233 Never instrument marked atomic accesses. This option can be used for
234 additional filtering. Conflicting marked atomic reads and plain
235 writes will never be reported as a data race, however, will cause
236 plain reads and marked writes to result in "unknown origin" reports.
237 If combined with CONFIG_KCSAN_REPORT_RACE_UNKNOWN_ORIGIN=n, data
238 races where at least one access is marked atomic will never be
239 reported.
240
241 Similar to KCSAN_ASSUME_PLAIN_WRITES_ATOMIC, but including unaligned
242 accesses, conflicting marked atomic reads and plain writes will not
243 be reported as data races; however, unlike that option, data races
244 due to two conflicting plain writes will be reported (aligned and
245 unaligned, if CONFIG_KCSAN_ASSUME_PLAIN_WRITES_ATOMIC=n).
246
247 config KCSAN_PERMISSIVE
248 bool "Enable all additional permissive rules"
249 depends on KCSAN_REPORT_VALUE_CHANGE_ONLY
250 help
251 Enable additional permissive rules to ignore certain classes of data
252 races (also see kernel/kcsan/permissive.h). None of the permissive
253 rules imply that such data races are generally safe, but can be used
254 to further reduce reported data races due to data-racy patterns
255 common across the kernel.
256
257 endif # KCSAN
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.