linux/lib/Kconfig.kcsan

258 lines
10 KiB
Plaintext
Raw Normal View History

# SPDX-License-Identifier: GPL-2.0-only
config HAVE_ARCH_KCSAN
bool
kcsan: Restrict supported compilers The first version of Clang that supports -tsan-distinguish-volatile will be able to support KCSAN. The first Clang release to do so, will be Clang 11. This is due to satisfying all the following requirements: 1. Never emit calls to __tsan_func_{entry,exit}. 2. __no_kcsan functions should not call anything, not even kcsan_{enable,disable}_current(), when using __{READ,WRITE}_ONCE => Requires leaving them plain! 3. Support atomic_{read,set}*() with KCSAN, which rely on arch_atomic_{read,set}*() using __{READ,WRITE}_ONCE() => Because of #2, rely on Clang 11's -tsan-distinguish-volatile support. We will double-instrument atomic_{read,set}*(), but that's reasonable given it's still lower cost than the data_race() variant due to avoiding 2 extra calls (kcsan_{en,dis}able_current() calls). 4. __always_inline functions inlined into __no_kcsan functions are never instrumented. 5. __always_inline functions inlined into instrumented functions are instrumented. 6. __no_kcsan_or_inline functions may be inlined into __no_kcsan functions => Implies leaving 'noinline' off of __no_kcsan_or_inline. 7. Because of #6, __no_kcsan and __no_kcsan_or_inline functions should never be spuriously inlined into instrumented functions, causing the accesses of the __no_kcsan function to be instrumented. Older versions of Clang do not satisfy #3. The latest GCC currently doesn't support at least #1, #3, and #7. Signed-off-by: Marco Elver <elver@google.com> Signed-off-by: Borislav Petkov <bp@suse.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Will Deacon <will@kernel.org> Link: https://lkml.kernel.org/r/CANpmjNMTsY_8241bS7=XAfqvZHFLrVEkv_uM4aDUWE_kh3Rvbw@mail.gmail.com Link: https://lkml.kernel.org/r/20200521142047.169334-7-elver@google.com
2020-05-21 14:20:42 +00:00
config HAVE_KCSAN_COMPILER
def_bool (CC_IS_CLANG && $(cc-option,-fsanitize=thread -mllvm -tsan-distinguish-volatile=1)) || \
(CC_IS_GCC && $(cc-option,-fsanitize=thread --param tsan-distinguish-volatile=1))
kcsan: Restrict supported compilers The first version of Clang that supports -tsan-distinguish-volatile will be able to support KCSAN. The first Clang release to do so, will be Clang 11. This is due to satisfying all the following requirements: 1. Never emit calls to __tsan_func_{entry,exit}. 2. __no_kcsan functions should not call anything, not even kcsan_{enable,disable}_current(), when using __{READ,WRITE}_ONCE => Requires leaving them plain! 3. Support atomic_{read,set}*() with KCSAN, which rely on arch_atomic_{read,set}*() using __{READ,WRITE}_ONCE() => Because of #2, rely on Clang 11's -tsan-distinguish-volatile support. We will double-instrument atomic_{read,set}*(), but that's reasonable given it's still lower cost than the data_race() variant due to avoiding 2 extra calls (kcsan_{en,dis}able_current() calls). 4. __always_inline functions inlined into __no_kcsan functions are never instrumented. 5. __always_inline functions inlined into instrumented functions are instrumented. 6. __no_kcsan_or_inline functions may be inlined into __no_kcsan functions => Implies leaving 'noinline' off of __no_kcsan_or_inline. 7. Because of #6, __no_kcsan and __no_kcsan_or_inline functions should never be spuriously inlined into instrumented functions, causing the accesses of the __no_kcsan function to be instrumented. Older versions of Clang do not satisfy #3. The latest GCC currently doesn't support at least #1, #3, and #7. Signed-off-by: Marco Elver <elver@google.com> Signed-off-by: Borislav Petkov <bp@suse.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Will Deacon <will@kernel.org> Link: https://lkml.kernel.org/r/CANpmjNMTsY_8241bS7=XAfqvZHFLrVEkv_uM4aDUWE_kh3Rvbw@mail.gmail.com Link: https://lkml.kernel.org/r/20200521142047.169334-7-elver@google.com
2020-05-21 14:20:42 +00:00
help
For the list of compilers that support KCSAN, please see
<file:Documentation/dev-tools/kcsan.rst>.
menuconfig KCSAN
bool "KCSAN: dynamic data race detector"
kcsan: Restrict supported compilers The first version of Clang that supports -tsan-distinguish-volatile will be able to support KCSAN. The first Clang release to do so, will be Clang 11. This is due to satisfying all the following requirements: 1. Never emit calls to __tsan_func_{entry,exit}. 2. __no_kcsan functions should not call anything, not even kcsan_{enable,disable}_current(), when using __{READ,WRITE}_ONCE => Requires leaving them plain! 3. Support atomic_{read,set}*() with KCSAN, which rely on arch_atomic_{read,set}*() using __{READ,WRITE}_ONCE() => Because of #2, rely on Clang 11's -tsan-distinguish-volatile support. We will double-instrument atomic_{read,set}*(), but that's reasonable given it's still lower cost than the data_race() variant due to avoiding 2 extra calls (kcsan_{en,dis}able_current() calls). 4. __always_inline functions inlined into __no_kcsan functions are never instrumented. 5. __always_inline functions inlined into instrumented functions are instrumented. 6. __no_kcsan_or_inline functions may be inlined into __no_kcsan functions => Implies leaving 'noinline' off of __no_kcsan_or_inline. 7. Because of #6, __no_kcsan and __no_kcsan_or_inline functions should never be spuriously inlined into instrumented functions, causing the accesses of the __no_kcsan function to be instrumented. Older versions of Clang do not satisfy #3. The latest GCC currently doesn't support at least #1, #3, and #7. Signed-off-by: Marco Elver <elver@google.com> Signed-off-by: Borislav Petkov <bp@suse.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Will Deacon <will@kernel.org> Link: https://lkml.kernel.org/r/CANpmjNMTsY_8241bS7=XAfqvZHFLrVEkv_uM4aDUWE_kh3Rvbw@mail.gmail.com Link: https://lkml.kernel.org/r/20200521142047.169334-7-elver@google.com
2020-05-21 14:20:42 +00:00
depends on HAVE_ARCH_KCSAN && HAVE_KCSAN_COMPILER
depends on DEBUG_KERNEL && !KASAN
select CONSTRUCTORS
select STACKTRACE
help
The Kernel Concurrency Sanitizer (KCSAN) is a dynamic
data-race detector that relies on compile-time instrumentation.
KCSAN uses a watchpoint-based sampling approach to detect races.
While KCSAN's primary purpose is to detect data races, it
also provides assertions to check data access constraints.
These assertions can expose bugs that do not manifest as
data races.
See <file:Documentation/dev-tools/kcsan.rst> for more details.
if KCSAN
config CC_HAS_TSAN_COMPOUND_READ_BEFORE_WRITE
def_bool (CC_IS_CLANG && $(cc-option,-fsanitize=thread -mllvm -tsan-compound-read-before-write=1)) || \
(CC_IS_GCC && $(cc-option,-fsanitize=thread --param tsan-compound-read-before-write=1))
help
The compiler instruments plain compound read-write operations
differently (++, --, +=, -=, |=, &=, etc.), which allows KCSAN to
distinguish them from other plain accesses. This is currently
supported by Clang 12 or later.
config KCSAN_VERBOSE
bool "Show verbose reports with more information about system state"
depends on PROVE_LOCKING
help
If enabled, reports show more information about the system state that
may help better analyze and debug races. This includes held locks and
IRQ trace events.
While this option should generally be benign, we call into more
external functions on report generation; if a race report is
generated from any one of them, system stability may suffer due to
deadlocks or recursion. If in doubt, say N.
config KCSAN_SELFTEST
bool "Perform short selftests on boot"
default y
help
Run KCSAN selftests on boot. On test failure, causes the kernel to
panic. Recommended to be enabled, ensuring critical functionality
works as intended.
config KCSAN_KUNIT_TEST
tristate "KCSAN test for integrated runtime behaviour" if !KUNIT_ALL_TESTS
default KUNIT_ALL_TESTS
depends on TRACEPOINTS && KUNIT
select TORTURE_TEST
help
KCSAN test focusing on behaviour of the integrated runtime. Tests
various race scenarios, and verifies the reports generated to
console. Makes use of KUnit for test organization, and the Torture
framework for test thread control.
Each test case may run at least up to KCSAN_REPORT_ONCE_IN_MS
milliseconds. Test run duration may be optimized by building the
kernel and KCSAN test with KCSAN_REPORT_ONCE_IN_MS set to a lower
than default value.
Say Y here if you want the test to be built into the kernel and run
during boot; say M if you want the test to build as a module; say N
if you are unsure.
config KCSAN_EARLY_ENABLE
bool "Early enable during boot"
default y
help
If KCSAN should be enabled globally as soon as possible. KCSAN can
later be enabled/disabled via debugfs.
config KCSAN_NUM_WATCHPOINTS
int "Number of available watchpoints"
default 64
help
Total number of available watchpoints. An address range maps into a
specific watchpoint slot as specified in kernel/kcsan/encoding.h.
Although larger number of watchpoints may not be usable due to
limited number of CPUs, a larger value helps to improve performance
due to reducing cache-line contention. The chosen default is a
conservative value; we should almost never observe "no_capacity"
events (see /sys/kernel/debug/kcsan).
config KCSAN_UDELAY_TASK
int "Delay in microseconds (for tasks)"
default 80
help
For tasks, the microsecond delay after setting up a watchpoint.
config KCSAN_UDELAY_INTERRUPT
int "Delay in microseconds (for interrupts)"
default 20
help
For interrupts, the microsecond delay after setting up a watchpoint.
Interrupts have tighter latency requirements, and their delay should
be lower than for tasks.
config KCSAN_DELAY_RANDOMIZE
bool "Randomize above delays"
default y
help
If delays should be randomized, where the maximum is KCSAN_UDELAY_*.
If false, the chosen delays are always the KCSAN_UDELAY_* values
as defined above.
config KCSAN_SKIP_WATCH
int "Skip instructions before setting up watchpoint"
default 4000
help
The number of per-CPU memory operations to skip, before another
watchpoint is set up, i.e. one in KCSAN_SKIP_WATCH per-CPU
memory operations are used to set up a watchpoint. A smaller value
results in more aggressive race detection, whereas a larger value
improves system performance at the cost of missing some races.
config KCSAN_SKIP_WATCH_RANDOMIZE
bool "Randomize watchpoint instruction skip count"
default y
help
If instruction skip count should be randomized, where the maximum is
KCSAN_SKIP_WATCH. If false, the chosen value is always
KCSAN_SKIP_WATCH.
config KCSAN_INTERRUPT_WATCHER
bool "Interruptible watchers" if !KCSAN_STRICT
default KCSAN_STRICT
help
If enabled, a task that set up a watchpoint may be interrupted while
delayed. This option will allow KCSAN to detect races between
interrupted tasks and other threads of execution on the same CPU.
Currently disabled by default, because not all safe per-CPU access
primitives and patterns may be accounted for, and therefore could
result in false positives.
config KCSAN_REPORT_ONCE_IN_MS
int "Duration in milliseconds, in which any given race is only reported once"
default 3000
help
Any given race is only reported once in the defined time window.
Different races may still generate reports within a duration that is
smaller than the duration defined here. This allows rate limiting
reporting to avoid flooding the console with reports. Setting this
to 0 disables rate limiting.
# The main purpose of the below options is to control reported data races, and
# are not expected to be switched frequently by non-testers or at runtime.
# The defaults are chosen to be conservative, and can miss certain bugs.
config KCSAN_REPORT_RACE_UNKNOWN_ORIGIN
bool "Report races of unknown origin"
default y
help
If KCSAN should report races where only one access is known, and the
conflicting access is of unknown origin. This type of race is
reported if it was only possible to infer a race due to a data value
change while an access is being delayed on a watchpoint.
config KCSAN_STRICT
bool "Strict data-race checking"
help
KCSAN will report data races with the strictest possible rules, which
closely aligns with the rules defined by the Linux-kernel memory
consistency model (LKMM).
config KCSAN_WEAK_MEMORY
bool "Enable weak memory modeling to detect missing memory barriers"
default y
depends on KCSAN_STRICT
# We can either let objtool nop __tsan_func_{entry,exit}() and builtin
# atomics instrumentation in .noinstr.text, or use a compiler that can
# implement __no_kcsan to really remove all instrumentation.
depends on !ARCH_WANTS_NO_INSTR || HAVE_NOINSTR_HACK || \
CC_IS_GCC || CLANG_VERSION >= 140000
select OBJTOOL if HAVE_NOINSTR_HACK
help
Enable support for modeling a subset of weak memory, which allows
detecting a subset of data races due to missing memory barriers.
Depends on KCSAN_STRICT, because the options strengthening certain
plain accesses by default (depending on !KCSAN_STRICT) reduce the
ability to detect any data races invoving reordered accesses, in
particular reordered writes.
Weak memory modeling relies on additional instrumentation and may
affect performance.
config KCSAN_REPORT_VALUE_CHANGE_ONLY
bool "Only report races where watcher observed a data value change"
default y
depends on !KCSAN_STRICT
help
If enabled and a conflicting write is observed via a watchpoint, but
the data value of the memory location was observed to remain
unchanged, do not report the data race.
config KCSAN_ASSUME_PLAIN_WRITES_ATOMIC
bool "Assume that plain aligned writes up to word size are atomic"
default y
depends on !KCSAN_STRICT
help
Assume that plain aligned writes up to word size are atomic by
default, and also not subject to other unsafe compiler optimizations
resulting in data races. This will cause KCSAN to not report data
races due to conflicts where the only plain accesses are aligned
writes up to word size: conflicts between marked reads and plain
aligned writes up to word size will not be reported as data races;
notice that data races between two conflicting plain aligned writes
will also not be reported.
config KCSAN_IGNORE_ATOMICS
bool "Do not instrument marked atomic accesses"
depends on !KCSAN_STRICT
help
Never instrument marked atomic accesses. This option can be used for
additional filtering. Conflicting marked atomic reads and plain
writes will never be reported as a data race, however, will cause
plain reads and marked writes to result in "unknown origin" reports.
If combined with CONFIG_KCSAN_REPORT_RACE_UNKNOWN_ORIGIN=n, data
races where at least one access is marked atomic will never be
reported.
Similar to KCSAN_ASSUME_PLAIN_WRITES_ATOMIC, but including unaligned
accesses, conflicting marked atomic reads and plain writes will not
be reported as data races; however, unlike that option, data races
due to two conflicting plain writes will be reported (aligned and
unaligned, if CONFIG_KCSAN_ASSUME_PLAIN_WRITES_ATOMIC=n).
config KCSAN_PERMISSIVE
bool "Enable all additional permissive rules"
depends on KCSAN_REPORT_VALUE_CHANGE_ONLY
help
Enable additional permissive rules to ignore certain classes of data
races (also see kernel/kcsan/permissive.h). None of the permissive
rules imply that such data races are generally safe, but can be used
to further reduce reported data races due to data-racy patterns
common across the kernel.
endif # KCSAN