xtensa: enable KCSAN
Subsystems (2)
Max Filippov
maintainerPATCH 1/1 xtensa: enable KCSAN
Prefix arch-specific barrier macros with '__' to make use of instrumented generic macros.
Prefix arch-specific bitops with 'arch_' to make use of instrumented generic functions.
Provide stubs for 64-bit atomics when building with KCSAN. Disable KCSAN instrumentation in arch/xtensa/boot.
--- arch/xtensa/Kconfig | 1 + arch/xtensa/boot/lib/Makefile | 1 + arch/xtensa/include/asm/barrier.h | 6 ++-- arch/xtensa/include/asm/bitops.h | 10 +++--- arch/xtensa/lib/Makefile | 2 ++ arch/xtensa/lib/kcsan-stubs.c | 54 +++++++++++++++++++++++++++++++ 6 files changed, 67 insertions(+), 7 deletions(-) create mode 100644 arch/xtensa/lib/kcsan-stubs.c
arch/xtensa/Kconfig +1 -0
@@ -29,6 +29,7 @@ config XTENSA select HAVE_ARCH_AUDITSYSCALL select HAVE_ARCH_JUMP_LABEL if !XIP_KERNEL select HAVE_ARCH_KASAN if MMU && !XIP_KERNEL + select HAVE_ARCH_KCSAN select HAVE_ARCH_SECCOMP_FILTER select HAVE_ARCH_TRACEHOOK select HAVE_CONTEXT_TRACKING
arch/xtensa/boot/lib/Makefile +1 -0
@@ -16,6 +16,7 @@ CFLAGS_REMOVE_inffast.o = -pg endif KASAN_SANITIZE := n +KCSAN_SANITIZE := n CFLAGS_REMOVE_inflate.o += -fstack-protector -fstack-protector-strong CFLAGS_REMOVE_zmem.o += -fstack-protector -fstack-protector-strong
arch/xtensa/include/asm/barrier.h +3 -3
@@ -11,9 +11,9 @@#include <asm/core.h> -#define mb () ({ __asm__ __volatile__("memw" : : : "memory"); }) -#define rmb () barrier() -#define wmb () mb() +#define __mb () ({ __asm__ __volatile__("memw" : : : "memory"); }) +#define __rmb () barrier() +#define __wmb () mb()#if XCHAL_HAVE_S32C1I #define __smp_mb__before_atomic () barrier()
arch/xtensa/include/asm/bitops.h +6 -4
@@ -99,7 +99,7 @@ static inline unsigned long __fls(unsigned long word)#if XCHAL_HAVE_EXCLUSIVE #define BIT_OP (op, insn, inv) \ -static inline void op##_bit(unsigned int bit, volatile unsigned long *p)\ +static inline void arch_##op##_bit(unsigned int bit, volatile unsigned long *p)\ { \ unsigned long tmp; \ unsigned long mask = 1UL << (bit & 31); \
@@ -119,7 +119,7 @@ static inline void op##_bit(unsigned int bit, volatile unsigned long *p)\#define TEST_AND_BIT_OP (op, insn, inv) \ static inline int \ -test_and_##op##_bit(unsigned int bit, volatile unsigned long *p) \ +arch_test_and_##op##_bit(unsigned int bit, volatile unsigned long *p) \ { \ unsigned long tmp, value; \ unsigned long mask = 1UL << (bit & 31); \
@@ -142,7 +142,7 @@ test_and_##op##_bit(unsigned int bit, volatile unsigned long *p) \#elif XCHAL_HAVE_S32C1I #define BIT_OP (op, insn, inv) \ -static inline void op##_bit(unsigned int bit, volatile unsigned long *p)\ +static inline void arch_##op##_bit(unsigned int bit, volatile unsigned long *p)\ { \ unsigned long tmp, value; \ unsigned long mask = 1UL << (bit & 31); \
@@ -163,7 +163,7 @@ static inline void op##_bit(unsigned int bit, volatile unsigned long *p)\#define TEST_AND_BIT_OP (op, insn, inv) \ static inline int \ -test_and_##op##_bit(unsigned int bit, volatile unsigned long *p) \ +arch_test_and_##op##_bit(unsigned int bit, volatile unsigned long *p) \ { \ unsigned long tmp, value; \ unsigned long mask = 1UL << (bit & 31); \
@@ -205,6 +205,8 @@ BIT_OPS(change, "xor", )#undef BIT_OP#undef TEST_AND_BIT_OP +#include <asm-generic/bitops/instrumented-atomic.h> +#include <asm-generic/bitops/le.h> #include <asm-generic/bitops/ext2-atomic-setbit.h>
arch/xtensa/lib/Makefile +2 -0
@@ -8,3 +8,5 @@ lib-y += memcopy.o memset.o checksum.o \ divsi3.o udivsi3.o modsi3.o umodsi3.o mulsi3.o \ usercopy.o strncpy_user.o strnlen_user.o lib-$(CONFIG_PCI) += pci-auto.o +lib-$(CONFIG_KCSAN) += kcsan-stubs.o +KCSAN_SANITIZE_kcsan-stubs.o := n
arch/xtensa/lib/kcsan-stubs.c new file +54 -0
54 lines not on this page · raw message
Signature · 1 line
2.30.2
Nice to see this happen!
Quoted 7 lines · Max Filippov
Prefix arch-specific barrier macros with '__' to make use of instrumented generic macros.
Prefix arch-specific bitops with 'arch_' to make use of instrumented generic functions.Provide stubs for 64-bit atomics when building with KCSAN.
The stubs are the only thing I don't understand. More elaboration on why this is required would be useful (maybe there's another way to solve?).
Quoted 1 line
Disable KCSAN instrumentation in
arch/xtensa/boot.
Given you went for barrier instrumentation, I assume you tested with a CONFIG_KCSAN_STRICT=y config? Did the kcsan_test pass?
Quoted 178 lines
Signed-off-by Max Filippov <[email protected]>--- arch/xtensa/Kconfig | 1 + arch/xtensa/boot/lib/Makefile | 1 + arch/xtensa/include/asm/barrier.h | 6 ++-- arch/xtensa/include/asm/bitops.h | 10 +++--- arch/xtensa/lib/Makefile | 2 ++ arch/xtensa/lib/kcsan-stubs.c | 54 +++++++++++++++++++++++++++++++ 6 files changed, 67 insertions(+), 7 deletions(-) create mode 100644 arch/xtensa/lib/kcsan-stubs.carch/xtensa/Kconfig +1
-0@@ -29,6 +29,7 @@ config XTENSA select HAVE_ARCH_AUDITSYSCALL select HAVE_ARCH_JUMP_LABEL if !XIP_KERNEL select HAVE_ARCH_KASAN if MMU && !XIP_KERNEL + select HAVE_ARCH_KCSAN select HAVE_ARCH_SECCOMP_FILTER select HAVE_ARCH_TRACEHOOK select HAVE_CONTEXT_TRACKINGarch/xtensa/boot/lib/Makefile +1
-0@@ -16,6 +16,7 @@ CFLAGS_REMOVE_inffast.o = -pg endif KASAN_SANITIZE := n +KCSAN_SANITIZE := n CFLAGS_REMOVE_inflate.o += -fstack-protector -fstack-protector-strong CFLAGS_REMOVE_zmem.o += -fstack-protector -fstack-protector-strongarch/xtensa/include/asm/barrier.h +3
-3@@ -11,9 +11,9 @@#include <asm/core.h> -#define mb () ({ __asm__ __volatile__("memw" : : : "memory"); }) -#define rmb () barrier() -#define wmb () mb() +#define __mb () ({ __asm__ __volatile__("memw" : : : "memory"); }) +#define __rmb () barrier() +#define __wmb () mb()#if XCHAL_HAVE_S32C1I #define __smp_mb__before_atomic () barrier()arch/xtensa/include/asm/bitops.h +6
-4@@ -99,7 +99,7 @@ static inline unsigned long __fls(unsigned long word)#if XCHAL_HAVE_EXCLUSIVE #define BIT_OP (op, insn, inv) \ -static inline void op##_bit(unsigned int bit, volatile unsigned long *p)\ +static inline void arch_##op##_bit(unsigned int bit, volatile unsigned long *p)\ { \ unsigned long tmp; \ unsigned long mask = 1UL << (bit & 31); \@@ -119,7 +119,7 @@ static inline void op##_bit(unsigned int bit, volatile unsigned long *p)\#define TEST_AND_BIT_OP (op, insn, inv) \ static inline int \ -test_and_##op##_bit(unsigned int bit, volatile unsigned long *p) \ +arch_test_and_##op##_bit(unsigned int bit, volatile unsigned long *p) \ { \ unsigned long tmp, value; \ unsigned long mask = 1UL << (bit & 31); \@@ -142,7 +142,7 @@ test_and_##op##_bit(unsigned int bit, volatile unsigned long *p) \#elif XCHAL_HAVE_S32C1I #define BIT_OP (op, insn, inv) \ -static inline void op##_bit(unsigned int bit, volatile unsigned long *p)\ +static inline void arch_##op##_bit(unsigned int bit, volatile unsigned long *p)\ { \ unsigned long tmp, value; \ unsigned long mask = 1UL << (bit & 31); \@@ -163,7 +163,7 @@ static inline void op##_bit(unsigned int bit, volatile unsigned long *p)\#define TEST_AND_BIT_OP (op, insn, inv) \ static inline int \ -test_and_##op##_bit(unsigned int bit, volatile unsigned long *p) \ +arch_test_and_##op##_bit(unsigned int bit, volatile unsigned long *p) \ { \ unsigned long tmp, value; \ unsigned long mask = 1UL << (bit & 31); \@@ -205,6 +205,8 @@ BIT_OPS(change, "xor", )#undef BIT_OP#undef TEST_AND_BIT_OP +#include <asm-generic/bitops/instrumented-atomic.h> +#include <asm-generic/bitops/le.h> #include <asm-generic/bitops/ext2-atomic-setbit.h> arch/xtensa/lib/Makefile +2
-0@@ -8,3 +8,5 @@ lib-y += memcopy.o memset.o checksum.o \ divsi3.o udivsi3.o modsi3.o umodsi3.o mulsi3.o \ usercopy.o strncpy_user.o strnlen_user.o lib-$(CONFIG_PCI) += pci-auto.o +lib-$(CONFIG_KCSAN) += kcsan-stubs.o +KCSAN_SANITIZE_kcsan-stubs.o := narch/xtensa/lib/kcsan-stubs.c new file +54
-1@@ -0,0 +1,54 @@ +// SPDX-License-Identifier: GPL-2.0 + +#include <linux/bug.h> +#include <linux/types.h> + +void __atomic_store_8 (volatile void *p,u64 v,int i) +{ +BUG (); +} + +u64 __atomic_load_8 (const volatile void *p,int i) +{ +BUG (); +} + +u64 __atomic_exchange_8 (volatile void *p,u64 v,int i) +{ +BUG (); +} + +bool __atomic_compare_exchange_8 (volatile void *p1,void *p2,u64 v,bool b,int i1,int i2) +{ +BUG (); +} + +u64 __atomic_fetch_add_8 (volatile void *p,u64 v,int i) +{ +BUG (); +} + +u64 __atomic_fetch_sub_8 (volatile void *p,u64 v,int i) +{ +BUG (); +} + +u64 __atomic_fetch_and_8 (volatile void *p,u64 v,int i) +{ +BUG (); +} + +u64 __atomic_fetch_or_8 (volatile void *p,u64 v,int i) +{ +BUG (); +} + +u64 __atomic_fetch_xor_8 (volatile void *p,u64 v,int i) +{ +BUG (); +} + +u64 __atomic_fetch_nand_8 (volatile void *p,u64 v,int i) +{ +BUG (); +} --2.30.2
--
You received this message because you are subscribed to the Google Groups "kasan-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/kasan-dev/20220416081355.2155050-1-jcmvbkbc%40gmail.com.
Max Filippov
maintainerHi Marco,
Quoted 9 lines · Marco Elver
Nice to see this happen!
Quoted 1 line · Max Filippov
Provide stubs for 64-bit atomics when building with KCSAN.
The stubs are the only thing I don't understand. More elaboration on why this is required would be useful (maybe there's another way to solve?).
It doesn't build without it, because the compiler left function calls in the code:
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.o: in function `__tsan_atomic32_compare_exchange_val':kernel/kcsan/core.c:1262: undefined reference to `__atomic_load_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.o: in function `__tsan_atomic64_load':kernel/kcsan/core.c:1262: undefined reference to `__atomic_load_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.c:1262: undefined reference to `__atomic_store_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.o: in function `__tsan_atomic64_store':kernel/kcsan/core.c:1262: undefined reference to `__atomic_store_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.c:1262: undefined reference to `__atomic_exchange_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.o: in function `__tsan_atomic64_exchange':kernel/kcsan/core.c:1262: undefined reference to `__atomic_exchange_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.c:1262: undefined reference to `__atomic_fetch_add_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.o: in function `__tsan_atomic64_fetch_add':kernel/kcsan/core.c:1262: undefined reference to `__atomic_fetch_add_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.c:1262: undefined reference to `__atomic_fetch_sub_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.o: in function `__tsan_atomic64_fetch_sub':kernel/kcsan/core.c:1262: undefined reference to `__atomic_fetch_sub_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.c:1262: undefined reference to `__atomic_fetch_and_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.o: in function `__tsan_atomic64_fetch_and':kernel/kcsan/core.c:1262: undefined reference to `__atomic_fetch_and_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.c:1262: undefined reference to `__atomic_fetch_or_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.o: in function `__tsan_atomic64_fetch_or':kernel/kcsan/core.c:1262: undefined reference to `__atomic_fetch_or_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.c:1262: undefined reference to `__atomic_fetch_xor_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.o: in function `__tsan_atomic64_fetch_xor':kernel/kcsan/core.c:1262: undefined reference to `__atomic_fetch_xor_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.c:1262: undefined reference to `__atomic_fetch_nand_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.o: in function `__tsan_atomic64_fetch_nand':kernel/kcsan/core.c:1262: undefined reference to `__atomic_fetch_nand_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.c:1262: undefined reference to `__atomic_compare_exchange_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.o: in function `__tsan_atomic64_compare_exchange_strong':kernel/kcsan/core.c:1262: undefined reference to `__atomic_compare_exchange_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.c:1262: undefined reference to `__atomic_compare_exchange_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.o: in function `__tsan_atomic64_compare_exchange_weak':kernel/kcsan/core.c:1262: undefined reference to `__atomic_compare_exchange_8'
xtensa-de233_fpu-elf-ld: kernel/kcsan/core.c:1262: undefined reference to `__atomic_compare_exchange_8'
None of these functions are called because xtensa doesn't have 64-bit atomic ops.
I guess that another way to fix it would be making DEFINE_TSAN_ATOMIC_OPS(64);
conditional and not enabling it when building for xtensa.
Quoted 4 lines
Quoted 1 line
Disable KCSAN instrumentation in
arch/xtensa/boot.Given you went for barrier instrumentation, I assume you tested with a
CONFIG_KCSAN_STRICT=yconfig?
Yes.
Quoted 1 line
Did the
kcsan_testpass?
current results are the following on QEMU:
# test_missing_barrier: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1313
Expected match_expect to be true, but is false
# test_atomic_builtins_missing_barrier: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1356
Expected match_expect to be true, but is false
# kcsan: pass:27 fail:2 skip:0 total:29
# Totals: pass:193 fail:4 skip:0 total:197
and the following on the real hardware:
# test_concurrent_races: EXPECTATION FAILED at kernel/kcsan/kcsan_test.c:762 Expected match_expect to be true, but is false
# test_write_write_struct_part: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:910
Expected match_expect to be true, but is false
# test_assert_exclusive_access_writer: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1077
Expected match_expect_access_writer to be true, but is false
# test_assert_exclusive_bits_change: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1098
Expected match_expect to be true, but is false
# test_assert_exclusive_writer_scoped: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1136
Expected match_expect_start to be true, but is false
# test_missing_barrier: EXPECTATION FAILED at kernel/kcsan/kcsan_test.c:1313 Expected match_expect to be true, but is false
# test_atomic_builtins_missing_barrier: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1356
Expected match_expect to be true, but is false
# kcsan: pass:22 fail:7 skip:0 total:29
# Totals: pass:177 fail:20 skip:0 total:197
Signature · 2 lines
Thanks. -- Max
On Tue, Apr 19, 2022 at 07:59PM -0700, Max Filippov wrote:
[...]
Quoted 12 lines
Quoted 3 lines
The stubs are the only thing I don't understand. More elaboration on why this is required would be useful (maybe there's another way to solve?).
It doesn't build without it, because the compiler left function calls in the code:
xtensa-de233_fpu-elf-ld:
kernel/kcsan/core.o: in function `__tsan_atomic32_compare_exchange_val':kernel/kcsan/core.c:1262: undefined reference to `__atomic_load_8'
xtensa-de233_fpu-elf-ld:kernel/kcsan/core.o: in function `__tsan_atomic64_load':
[...]
Quoted 7 lines
None of these functions are called because xtensa doesn't have 64-bit atomic ops.
I guess that another way to fix it would be making
DEFINE_TSAN_ATOMIC_OPS(64);
conditional and not enabling it when building for xtensa.
I see - however, it seems the kernel provides 64-bit atomics to xtensa using lib/atomic64.c:
arch/xtensa/Kconfig: select GENERIC_ATOMIC64
So the right thing to do might be to implement the builtin atomics using the kernel's atomic64_* primitives. However, granted, the builtin atomics might not be needed on xtensa (depending on configuration). Their existence is due to some compiler instrumentation emitting builtin-atomics (Clang's GCOV), folks using them accidentally and blaming KCSAN (also https://paulmck.livejournal.com/64970.html).
So I think it's fair to leave them to BUG() until somebody complains (at which point they need to be implemented). I leave it to you.
Quoted 35 lines
Quoted 4 lines
Quoted 1 line
Disable KCSAN instrumentation in
arch/xtensa/boot.Given you went for barrier instrumentation, I assume you tested with a
CONFIG_KCSAN_STRICT=yconfig?Yes.
Quoted 1 line
Did the
kcsan_testpass?current results are the following on QEMU:
#
test_missing_barrier: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1313
Expectedmatch_expectto be true, but is false
#test_atomic_builtins_missing_barrier: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1356
Expectedmatch_expectto be true, but is false
# kcsan: pass:27 fail:2 skip:0 total:29
# Totals: pass:193 fail:4 skip:0 total:197and the following on the real hardware:
#
test_concurrent_races: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:762Expectedmatch_expectto be true, but is false
#test_write_write_struct_part: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:910
Expectedmatch_expectto be true, but is false
#test_assert_exclusive_access_writer: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1077
Expectedmatch_expect_access_writerto be true, but is false
#test_assert_exclusive_bits_change: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1098
Expectedmatch_expectto be true, but is false
#test_assert_exclusive_writer_scoped: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1136
Quoted 8 lines
Expected
match_expect_startto be true, but is false
#test_missing_barrier: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1313Expectedmatch_expectto be true, but is false
#test_atomic_builtins_missing_barrier: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1356
Expectedmatch_expectto be true, but is false
# kcsan: pass:22 fail:7 skip:0 total:29
# Totals: pass:177 fail:20 skip:0 total:197
Each test case is run with varying number of threads - am I correctly inferring that out of all test cases, usually only one such run failed, and runs with different number of threads (of the same test case) succeeded?
If that's the case, I think we can say that it works, and the failures are due to flakiness with either higher or lower threads counts. I know that some test cases might still be flaky under QEMU TCG because of how it does concurrent execution of different CPU cores.
Thanks,
-- Marco
Max Filippov
maintainerQuoted 9 lines · Marco Elver
So the right thing to do might be to implement the builtin atomics using the kernel's atomic64_* primitives. However, granted, the builtin atomics might not be needed on xtensa (depending on configuration). Their existence is due to some compiler instrumentation emitting builtin-atomics (Clang's GCOV), folks using them accidentally and blaming KCSAN (also https://paulmck.livejournal.com/64970.html).
So I think it's fair to leave them to
BUG()until somebody complains (at which point they need to be implemented). I leave it to you.
Sure, that was my plan.
Quoted 37 lines
Quoted 28 lines
Quoted 1 line
Did the
kcsan_testpass?current results are the following on QEMU:
#
test_missing_barrier: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1313
Expectedmatch_expectto be true, but is false
#test_atomic_builtins_missing_barrier: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1356
Expectedmatch_expectto be true, but is false
# kcsan: pass:27 fail:2 skip:0 total:29
# Totals: pass:193 fail:4 skip:0 total:197and the following on the real hardware:
#
test_concurrent_races: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:762Expectedmatch_expectto be true, but is false
#test_write_write_struct_part: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:910
Expectedmatch_expectto be true, but is false
#test_assert_exclusive_access_writer: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1077
Expectedmatch_expect_access_writerto be true, but is false
#test_assert_exclusive_bits_change: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1098
Expectedmatch_expectto be true, but is false
#test_assert_exclusive_writer_scoped: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1136Quoted 8 lines
Expected
match_expect_startto be true, but is false
#test_missing_barrier: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1313Expectedmatch_expectto be true, but is false
#test_atomic_builtins_missing_barrier: EXPECTATION FAILED atkernel/kcsan/kcsan_test.c:1356
Expectedmatch_expectto be true, but is false
# kcsan: pass:22 fail:7 skip:0 total:29
# Totals: pass:177 fail:20 skip:0 total:197
Quoted 4 lines
Each test case is run with varying number of threads - am I correctly inferring that out of all test cases, usually only one such run failed, and runs with different number of threads (of the same test case) succeeded?
For most of the failures -- yes.
For the test_missing_barrier and test_atomic_builtins_missing_barrier on the hardware it was the opposite: only one subtest succeeded while all others failed. Does it mean that the xtensa memory model is insufficiently weak?
Quoted 4 lines
If that's the case, I think we can say that it works, and the failures are due to flakiness with either higher or lower threads counts. I know that some test cases might still be flaky under QEMU TCG because of how it does concurrent execution of different CPU cores.
Thanks for taking a look.
I'll post v2 with a couple additional minor changes.
Signature · 2 lines
Thanks. -- Max
Quoted 10 lines · Max Filippov
Quoted 4 lines
Each test case is run with varying number of threads - am I correctly inferring that out of all test cases, usually only one such run failed, and runs with different number of threads (of the same test case) succeeded?
For most of the failures -- yes.
For thetest_missing_barrierandtest_atomic_builtins_missing_barrieron the hardware it was the opposite: only one subtest succeeded while all others failed. Does it mean that the xtensa memory model is insufficiently weak?
No - KCSAN's weak memory modeling and detection of missing barriers doesn't care what the HW does, it only approximates the LKMM. If the test_barrier_nothreads case passed, there's nothing wrong with barrier instrumentation. Regarding the test case failures, if at least 1 passed I'm guessing it's just flaky (not enough concurrency, or unexpected barriers due to too many interrupts which can happen if we enter the scheduler).
Unfortunately I don't know xtensa and if this is normal, but modulo flakiness, I think it's fine. (I've made a note to try and deflake the test if I can find time to try the xtensa version.)