1# SPDX-License-Identifier: GPL-2.0-only 2 3config HAVE_ARCH_KCSAN 4 bool 5 6config HAVE_KCSAN_COMPILER 7 def_bool CC_IS_CLANG && $(cc-option,-fsanitize=thread -mllvm -tsan-distinguish-volatile=1) 8 help 9 For the list of compilers that support KCSAN, please see 10 <file:Documentation/dev-tools/kcsan.rst>. 11 12config KCSAN_KCOV_BROKEN 13 def_bool KCOV && CC_HAS_SANCOV_TRACE_PC 14 depends on CC_IS_CLANG 15 depends on !$(cc-option,-Werror=unused-command-line-argument -fsanitize=thread -fsanitize-coverage=trace-pc) 16 help 17 Some versions of clang support either KCSAN and KCOV but not the 18 combination of the two. 19 See https://bugs.llvm.org/show_bug.cgi?id=45831 for the status 20 in newer releases. 21 22menuconfig KCSAN 23 bool "KCSAN: dynamic data race detector" 24 depends on HAVE_ARCH_KCSAN && HAVE_KCSAN_COMPILER 25 depends on DEBUG_KERNEL && !KASAN 26 depends on !KCSAN_KCOV_BROKEN 27 select STACKTRACE 28 help 29 The Kernel Concurrency Sanitizer (KCSAN) is a dynamic 30 data-race detector that relies on compile-time instrumentation. 31 KCSAN uses a watchpoint-based sampling approach to detect races. 32 33 While KCSAN's primary purpose is to detect data races, it 34 also provides assertions to check data access constraints. 35 These assertions can expose bugs that do not manifest as 36 data races. 37 38 See <file:Documentation/dev-tools/kcsan.rst> for more details. 39 40if KCSAN 41 42config 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 55config KCSAN_DEBUG 56 bool "Debugging of KCSAN internals" 57 58config KCSAN_SELFTEST 59 bool "Perform short selftests on boot" 60 default y 61 help 62 Run KCSAN selftests on boot. On test failure, causes the kernel to panic. 63 64config KCSAN_EARLY_ENABLE 65 bool "Early enable during boot" 66 default y 67 help 68 If KCSAN should be enabled globally as soon as possible. KCSAN can 69 later be enabled/disabled via debugfs. 70 71config KCSAN_NUM_WATCHPOINTS 72 int "Number of available watchpoints" 73 default 64 74 help 75 Total number of available watchpoints. An address range maps into a 76 specific watchpoint slot as specified in kernel/kcsan/encoding.h. 77 Although larger number of watchpoints may not be usable due to 78 limited number of CPUs, a larger value helps to improve performance 79 due to reducing cache-line contention. The chosen default is a 80 conservative value; we should almost never observe "no_capacity" 81 events (see /sys/kernel/debug/kcsan). 82 83config KCSAN_UDELAY_TASK 84 int "Delay in microseconds (for tasks)" 85 default 80 86 help 87 For tasks, the microsecond delay after setting up a watchpoint. 88 89config KCSAN_UDELAY_INTERRUPT 90 int "Delay in microseconds (for interrupts)" 91 default 20 92 help 93 For interrupts, the microsecond delay after setting up a watchpoint. 94 Interrupts have tighter latency requirements, and their delay should 95 be lower than for tasks. 96 97config KCSAN_DELAY_RANDOMIZE 98 bool "Randomize above delays" 99 default y 100 help 101 If delays should be randomized, where the maximum is KCSAN_UDELAY_*. 102 If false, the chosen delays are always the KCSAN_UDELAY_* values 103 as defined above. 104 105config KCSAN_SKIP_WATCH 106 int "Skip instructions before setting up watchpoint" 107 default 4000 108 help 109 The number of per-CPU memory operations to skip, before another 110 watchpoint is set up, i.e. one in KCSAN_WATCH_SKIP per-CPU 111 memory operations are used to set up a watchpoint. A smaller value 112 results in more aggressive race detection, whereas a larger value 113 improves system performance at the cost of missing some races. 114 115config KCSAN_SKIP_WATCH_RANDOMIZE 116 bool "Randomize watchpoint instruction skip count" 117 default y 118 help 119 If instruction skip count should be randomized, where the maximum is 120 KCSAN_WATCH_SKIP. If false, the chosen value is always 121 KCSAN_WATCH_SKIP. 122 123config KCSAN_INTERRUPT_WATCHER 124 bool "Interruptible watchers" 125 help 126 If enabled, a task that set up a watchpoint may be interrupted while 127 delayed. This option will allow KCSAN to detect races between 128 interrupted tasks and other threads of execution on the same CPU. 129 130 Currently disabled by default, because not all safe per-CPU access 131 primitives and patterns may be accounted for, and therefore could 132 result in false positives. 133 134config KCSAN_REPORT_ONCE_IN_MS 135 int "Duration in milliseconds, in which any given race is only reported once" 136 default 3000 137 help 138 Any given race is only reported once in the defined time window. 139 Different races may still generate reports within a duration that is 140 smaller than the duration defined here. This allows rate limiting 141 reporting to avoid flooding the console with reports. Setting this 142 to 0 disables rate limiting. 143 144# The main purpose of the below options is to control reported data races (e.g. 145# in fuzzer configs), and are not expected to be switched frequently by other 146# users. We could turn some of them into boot parameters, but given they should 147# not be switched normally, let's keep them here to simplify configuration. 148# 149# The defaults below are chosen to be very conservative, and may miss certain 150# bugs. 151 152config KCSAN_REPORT_RACE_UNKNOWN_ORIGIN 153 bool "Report races of unknown origin" 154 default y 155 help 156 If KCSAN should report races where only one access is known, and the 157 conflicting access is of unknown origin. This type of race is 158 reported if it was only possible to infer a race due to a data value 159 change while an access is being delayed on a watchpoint. 160 161config KCSAN_REPORT_VALUE_CHANGE_ONLY 162 bool "Only report races where watcher observed a data value change" 163 default y 164 help 165 If enabled and a conflicting write is observed via a watchpoint, but 166 the data value of the memory location was observed to remain 167 unchanged, do not report the data race. 168 169config KCSAN_ASSUME_PLAIN_WRITES_ATOMIC 170 bool "Assume that plain aligned writes up to word size are atomic" 171 default y 172 help 173 Assume that plain aligned writes up to word size are atomic by 174 default, and also not subject to other unsafe compiler optimizations 175 resulting in data races. This will cause KCSAN to not report data 176 races due to conflicts where the only plain accesses are aligned 177 writes up to word size: conflicts between marked reads and plain 178 aligned writes up to word size will not be reported as data races; 179 notice that data races between two conflicting plain aligned writes 180 will also not be reported. 181 182config KCSAN_IGNORE_ATOMICS 183 bool "Do not instrument marked atomic accesses" 184 help 185 Never instrument marked atomic accesses. This option can be used for 186 additional filtering. Conflicting marked atomic reads and plain 187 writes will never be reported as a data race, however, will cause 188 plain reads and marked writes to result in "unknown origin" reports. 189 If combined with CONFIG_KCSAN_REPORT_RACE_UNKNOWN_ORIGIN=n, data 190 races where at least one access is marked atomic will never be 191 reported. 192 193 Similar to KCSAN_ASSUME_PLAIN_WRITES_ATOMIC, but including unaligned 194 accesses, conflicting marked atomic reads and plain writes will not 195 be reported as data races; however, unlike that option, data races 196 due to two conflicting plain writes will be reported (aligned and 197 unaligned, if CONFIG_KCSAN_ASSUME_PLAIN_WRITES_ATOMIC=n). 198 199endif # KCSAN 200