1 # SPDX-License-Identifier: GPL-2.0-only
6 config HAVE_KCSAN_COMPILER
7 def_bool (CC_IS_CLANG && $(cc-option,-fsanitize=thread -mllvm -tsan-distinguish-volatile=1)) || \
8 (CC_IS_GCC && $(cc-option,-fsanitize=thread --param tsan-distinguish-volatile=1))
10 For the list of compilers that support KCSAN, please see
11 <file:Documentation/dev-tools/kcsan.rst>.
13 config KCSAN_KCOV_BROKEN
14 def_bool KCOV && CC_HAS_SANCOV_TRACE_PC
15 depends on CC_IS_CLANG
16 depends on !$(cc-option,-Werror=unused-command-line-argument -fsanitize=thread -fsanitize-coverage=trace-pc)
18 Some versions of clang support either KCSAN and KCOV but not the
19 combination of the two.
20 See https://bugs.llvm.org/show_bug.cgi?id=45831 for the status
24 bool "KCSAN: dynamic data race detector"
25 depends on HAVE_ARCH_KCSAN && HAVE_KCSAN_COMPILER
26 depends on DEBUG_KERNEL && !KASAN
27 depends on !KCSAN_KCOV_BROKEN
30 The Kernel Concurrency Sanitizer (KCSAN) is a dynamic
31 data-race detector that relies on compile-time instrumentation.
32 KCSAN uses a watchpoint-based sampling approach to detect races.
34 While KCSAN's primary purpose is to detect data races, it
35 also provides assertions to check data access constraints.
36 These assertions can expose bugs that do not manifest as
39 See <file:Documentation/dev-tools/kcsan.rst> for more details.
44 bool "Show verbose reports with more information about system state"
45 depends on PROVE_LOCKING
47 If enabled, reports show more information about the system state that
48 may help better analyze and debug races. This includes held locks and
51 While this option should generally be benign, we call into more
52 external functions on report generation; if a race report is
53 generated from any one of them, system stability may suffer due to
54 deadlocks or recursion. If in doubt, say N.
57 bool "Debugging of KCSAN internals"
60 bool "Perform short selftests on boot"
63 Run KCSAN selftests on boot. On test failure, causes the kernel to
64 panic. Recommended to be enabled, ensuring critical functionality
68 tristate "KCSAN test for integrated runtime behaviour"
69 depends on TRACEPOINTS && KUNIT
72 KCSAN test focusing on behaviour of the integrated runtime. Tests
73 various race scenarios, and verifies the reports generated to
74 console. Makes use of KUnit for test organization, and the Torture
75 framework for test thread control.
77 Each test case may run at least up to KCSAN_REPORT_ONCE_IN_MS
78 milliseconds. Test run duration may be optimized by building the
79 kernel and KCSAN test with KCSAN_REPORT_ONCE_IN_MS set to a lower
82 Say Y here if you want the test to be built into the kernel and run
83 during boot; say M if you want the test to build as a module; say N
86 config KCSAN_EARLY_ENABLE
87 bool "Early enable during boot"
90 If KCSAN should be enabled globally as soon as possible. KCSAN can
91 later be enabled/disabled via debugfs.
93 config KCSAN_NUM_WATCHPOINTS
94 int "Number of available watchpoints"
97 Total number of available watchpoints. An address range maps into a
98 specific watchpoint slot as specified in kernel/kcsan/encoding.h.
99 Although larger number of watchpoints may not be usable due to
100 limited number of CPUs, a larger value helps to improve performance
101 due to reducing cache-line contention. The chosen default is a
102 conservative value; we should almost never observe "no_capacity"
103 events (see /sys/kernel/debug/kcsan).
105 config KCSAN_UDELAY_TASK
106 int "Delay in microseconds (for tasks)"
109 For tasks, the microsecond delay after setting up a watchpoint.
111 config KCSAN_UDELAY_INTERRUPT
112 int "Delay in microseconds (for interrupts)"
115 For interrupts, the microsecond delay after setting up a watchpoint.
116 Interrupts have tighter latency requirements, and their delay should
117 be lower than for tasks.
119 config KCSAN_DELAY_RANDOMIZE
120 bool "Randomize above delays"
123 If delays should be randomized, where the maximum is KCSAN_UDELAY_*.
124 If false, the chosen delays are always the KCSAN_UDELAY_* values
127 config KCSAN_SKIP_WATCH
128 int "Skip instructions before setting up watchpoint"
131 The number of per-CPU memory operations to skip, before another
132 watchpoint is set up, i.e. one in KCSAN_WATCH_SKIP per-CPU
133 memory operations are used to set up a watchpoint. A smaller value
134 results in more aggressive race detection, whereas a larger value
135 improves system performance at the cost of missing some races.
137 config KCSAN_SKIP_WATCH_RANDOMIZE
138 bool "Randomize watchpoint instruction skip count"
141 If instruction skip count should be randomized, where the maximum is
142 KCSAN_WATCH_SKIP. If false, the chosen value is always
145 config KCSAN_INTERRUPT_WATCHER
146 bool "Interruptible watchers"
148 If enabled, a task that set up a watchpoint may be interrupted while
149 delayed. This option will allow KCSAN to detect races between
150 interrupted tasks and other threads of execution on the same CPU.
152 Currently disabled by default, because not all safe per-CPU access
153 primitives and patterns may be accounted for, and therefore could
154 result in false positives.
156 config KCSAN_REPORT_ONCE_IN_MS
157 int "Duration in milliseconds, in which any given race is only reported once"
160 Any given race is only reported once in the defined time window.
161 Different races may still generate reports within a duration that is
162 smaller than the duration defined here. This allows rate limiting
163 reporting to avoid flooding the console with reports. Setting this
164 to 0 disables rate limiting.
166 # The main purpose of the below options is to control reported data races (e.g.
167 # in fuzzer configs), and are not expected to be switched frequently by other
168 # users. We could turn some of them into boot parameters, but given they should
169 # not be switched normally, let's keep them here to simplify configuration.
171 # The defaults below are chosen to be very conservative, and may miss certain
174 config KCSAN_REPORT_RACE_UNKNOWN_ORIGIN
175 bool "Report races of unknown origin"
178 If KCSAN should report races where only one access is known, and the
179 conflicting access is of unknown origin. This type of race is
180 reported if it was only possible to infer a race due to a data value
181 change while an access is being delayed on a watchpoint.
183 config KCSAN_REPORT_VALUE_CHANGE_ONLY
184 bool "Only report races where watcher observed a data value change"
187 If enabled and a conflicting write is observed via a watchpoint, but
188 the data value of the memory location was observed to remain
189 unchanged, do not report the data race.
191 config KCSAN_ASSUME_PLAIN_WRITES_ATOMIC
192 bool "Assume that plain aligned writes up to word size are atomic"
195 Assume that plain aligned writes up to word size are atomic by
196 default, and also not subject to other unsafe compiler optimizations
197 resulting in data races. This will cause KCSAN to not report data
198 races due to conflicts where the only plain accesses are aligned
199 writes up to word size: conflicts between marked reads and plain
200 aligned writes up to word size will not be reported as data races;
201 notice that data races between two conflicting plain aligned writes
202 will also not be reported.
204 config KCSAN_IGNORE_ATOMICS
205 bool "Do not instrument marked atomic accesses"
207 Never instrument marked atomic accesses. This option can be used for
208 additional filtering. Conflicting marked atomic reads and plain
209 writes will never be reported as a data race, however, will cause
210 plain reads and marked writes to result in "unknown origin" reports.
211 If combined with CONFIG_KCSAN_REPORT_RACE_UNKNOWN_ORIGIN=n, data
212 races where at least one access is marked atomic will never be
215 Similar to KCSAN_ASSUME_PLAIN_WRITES_ATOMIC, but including unaligned
216 accesses, conflicting marked atomic reads and plain writes will not
217 be reported as data races; however, unlike that option, data races
218 due to two conflicting plain writes will be reported (aligned and
219 unaligned, if CONFIG_KCSAN_ASSUME_PLAIN_WRITES_ATOMIC=n).