linux.git/tools/testing/selftests/futex/include, branch master Linux kernel source tree selftests/futex: Provide thread creation and synchronization helpers 2026-07-20T18:58:51+00:00 Yuwen Chen ywen.chen@foxmail.com 2026-05-18T02:16:40+00:00 6c4a1b972643d72bd26f674677df8b5a6a3fe94f There are timing issues in the use of threads in some selftests for futexes as the tests rely on timed waits to ensure that the other test thread[s] reached the lock wait function in the kernel. That "works" on halfways idle systems, but fails under load which results in tests failing. Provide a set of helper functions to create test threads and to wait for them to reach the lock wait by monitoring /proc/$PID/wchan. [ tglx: Fixup coding style, move the timeout into the helper, adapt to test harness changes and massage change log ] Signed-off-by: Yuwen Chen <ywen.chen@foxmail.com> Signed-off-by: Thomas Gleixner <tglx@kernel.org> Link: https://patch.msgid.link/tencent_4C90BE98BBC48B38B9FBC6A95C45CF49F309@qq.com 
There are timing issues in the use of threads in some selftests for futexes
as the tests rely on timed waits to ensure that the other test thread[s]
reached the lock wait function in the kernel.

That "works" on halfways idle systems, but fails under load which results
in tests failing.

Provide a set of helper functions to create test threads and to wait for
them to reach the lock wait by monitoring /proc/$PID/wchan.

[ tglx: Fixup coding style, move the timeout into the helper, adapt to test
  	harness changes and massage change log ]

Signed-off-by: Yuwen Chen <ywen.chen@foxmail.com>
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Link: https://patch.msgid.link/tencent_4C90BE98BBC48B38B9FBC6A95C45CF49F309@qq.com
selftests/futex: Migrate functional tests to harness 2026-07-05T19:49:17+00:00 Wake Liu wakel@google.com 2026-05-26T01:06:34+00:00 e531301dd8fa23f1edeee9c8af62310a4e1424c5 Currently, multiple futex functional tests (wait_timeout, waitv, wait_wouldblock) mix low-level ksft_* logging and result APIs with the kselftest_harness.h framework. On older kernels where system calls like futex_waitv are missing (returning -ENOSYS), this mixed usage triggers framework inconsistencies, causing the test to fail with: "Illegal usage of low-level ksft APIs in harness test". Address this by completely refactoring these tests to exclusively use the high-level kselftest_harness.h framework (gtest-like API), mapping all low-level calls to native harness macros: - Non-fatal assertions: Replace ksft_test_result_fail() with EXPECT_EQ() and EXPECT_NE(). - Fatal assertions: Replace ksft_exit_fail_msg() with ASSERT_EQ() and ASSERT_TRUE() to immediately terminate execution upon setup failure. - Test skipping: Replace ksft_exit_skip() with the graceful SKIP() macro combined with early runtime availability checks for sys_futex_waitv. - Debug logging: Replace ksft_print_dbg_msg() with TH_LOG() to inherit automatic file/line context. - Success reporting: Remove explicit ksft_test_result_pass() calls, deferring to automated harness completion reporting. Additionally: - Fix a critical SIGSEGV crash in early syscall probing logic caused by passing a NULL pointer to inline user-space timespec conversions. - Introduce TEST_TIMEOUT() and GET_ABS_TIMEOUT() macros in wait_timeout to encapsulate assertions while preserving source line attribution. - Pass test metadata (_metadata) into secondary threads to allow native harness assertions during concurrent execution. [ tglx: Fixup coding style ] Signed-off-by: Wake Liu <wakel@google.com> Signed-off-by: Thomas Gleixner <tglx@kernel.org> Link: https://patch.msgid.link/20260526010635.23980-2-wakel@google.com 
Currently, multiple futex functional tests (wait_timeout, waitv,
wait_wouldblock) mix low-level ksft_* logging and result APIs with the
kselftest_harness.h framework. On older kernels where system calls like
futex_waitv are missing (returning -ENOSYS), this mixed usage triggers
framework inconsistencies, causing the test to fail with: "Illegal usage of
low-level ksft APIs in harness test".

Address this by completely refactoring these tests to exclusively use
the high-level kselftest_harness.h framework (gtest-like API), mapping
all low-level calls to native harness macros:

  - Non-fatal assertions: Replace ksft_test_result_fail() with EXPECT_EQ()
    and EXPECT_NE().

  - Fatal assertions: Replace ksft_exit_fail_msg() with ASSERT_EQ() and
    ASSERT_TRUE() to immediately terminate execution upon setup failure.

  - Test skipping: Replace ksft_exit_skip() with the graceful SKIP() macro
    combined with early runtime availability checks for sys_futex_waitv.

  - Debug logging: Replace ksft_print_dbg_msg() with TH_LOG() to inherit
    automatic file/line context.

  - Success reporting: Remove explicit ksft_test_result_pass() calls,
    deferring to automated harness completion reporting.

Additionally:

  - Fix a critical SIGSEGV crash in early syscall probing logic caused by
    passing a NULL pointer to inline user-space timespec conversions.

  - Introduce TEST_TIMEOUT() and GET_ABS_TIMEOUT() macros in wait_timeout
    to encapsulate assertions while preserving source line attribution.

  - Pass test metadata (_metadata) into secondary threads to allow native
    harness assertions during concurrent execution.

[ tglx: Fixup coding style ]

Signed-off-by: Wake Liu <wakel@google.com>
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Link: https://patch.msgid.link/20260526010635.23980-2-wakel@google.com
selftests: futex: Add tests for robust release operations 2026-06-03T09:38:53+00:00 André Almeida andrealmeid@igalia.com 2026-06-02T09:10:21+00:00 608323bf7bb85bbb647eca4373acef247f105e67 Add tests for __vdso_futex_robust_listXX_try_unlock() and for the futex() op FUTEX_ROBUST_UNLOCK. Test the contended and uncontended cases for the vDSO functions and all ops combinations for FUTEX_ROBUST_UNLOCK. [ tglx: Replace the VDSO function lookup ] Signed-off-by: André Almeida <andrealmeid@igalia.com> Signed-off-by: Thomas Gleixner <tglx@kernel.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://patch.msgid.link/20260329-tonyk-vdso_test-v2-2-b7db810e44a1@igalia.com Link: https://patch.msgid.link/20260602090535.988101541@kernel.org 
Add tests for __vdso_futex_robust_listXX_try_unlock() and for the futex()
op FUTEX_ROBUST_UNLOCK.

Test the contended and uncontended cases for the vDSO functions and all
ops combinations for FUTEX_ROBUST_UNLOCK.

[ tglx: Replace the VDSO function lookup ]

Signed-off-by: André Almeida <andrealmeid@igalia.com>
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20260329-tonyk-vdso_test-v2-2-b7db810e44a1@igalia.com
Link: https://patch.msgid.link/20260602090535.988101541@kernel.org
selftests/futex: Remove logging.h file 2025-09-20T16:11:56+00:00 André Almeida andrealmeid@igalia.com 2025-09-17T21:21:54+00:00 520db0559deff096c33a95dd3be2583e02771261 Every futex selftest uses the kselftest_harness.h helper and don't need the logging.h file. Delete it. Signed-off-by: André Almeida <andrealmeid@igalia.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> 
Every futex selftest uses the kselftest_harness.h helper and don't need
the logging.h file. Delete it.

Signed-off-by: André Almeida <andrealmeid@igalia.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
selftests/futex: Fix futex_wait() for 32bit ARM 2025-09-01T14:26:55+00:00 Dan Carpenter dan.carpenter@linaro.org 2025-08-27T13:00:11+00:00 237bfb76c90b184f57bb18fe35ff366c19393dc8 On 32bit ARM systems gcc-12 will use 32bit timestamps while gcc-13 and later will use 64bit timestamps. The problem is that SYS_futex will continue pointing at the 32bit system call. This makes the futex_wait test fail like this: waiter failed errno 110 not ok 1 futex_wake private returned: 0 Success waiter failed errno 110 not ok 2 futex_wake shared (page anon) returned: 0 Success waiter failed errno 110 not ok 3 futex_wake shared (file backed) returned: 0 Success Instead of compiling differently depending on the gcc version, use the -D_FILE_OFFSET_BITS=64 -D_TIME_BITS=64 options to ensure that 64bit timestamps are used. Then use ifdefs to make SYS_futex point to the 64bit system call. Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Reviewed-by: André Almeida <andrealmeid@igalia.com> Tested-by: Anders Roxell <anders.roxell@linaro.org> Link: https://lore.kernel.org/20250827130011.677600-6-bigeasy@linutronix.de 
On 32bit ARM systems gcc-12 will use 32bit timestamps while gcc-13 and later
will use 64bit timestamps.  The problem is that SYS_futex will continue
pointing at the 32bit system call.  This makes the futex_wait test fail like
this:

  waiter failed errno 110
  not ok 1 futex_wake private returned: 0 Success
  waiter failed errno 110
  not ok 2 futex_wake shared (page anon) returned: 0 Success
  waiter failed errno 110
  not ok 3 futex_wake shared (file backed) returned: 0 Success

Instead of compiling differently depending on the gcc version, use the
-D_FILE_OFFSET_BITS=64 -D_TIME_BITS=64 options to ensure that 64bit timestamps
are used.  Then use ifdefs to make SYS_futex point to the 64bit system call.

Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: André Almeida <andrealmeid@igalia.com>
Tested-by: Anders Roxell <anders.roxell@linaro.org>
Link: https://lore.kernel.org/20250827130011.677600-6-bigeasy@linutronix.de
Merge tag 'locking-futex-2025-07-29' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip 2025-07-29T21:39:42+00:00 Linus Torvalds torvalds@linux-foundation.org 2025-07-29T21:39:42+00:00 f38b1f243ec3babea9d8d9c6240249589853aca2 Pull futex updates from Thomas Gleixner: - Switch the reference counting to a RCU based per-CPU reference to address a performance bottleneck vs the single instance rcuref variant - Make the futex selftest build on 32-bit architectures which only support 64-bit time_t, e.g. RISCV-32 - Cleanups and improvements in selftests and futex bench * tag 'locking-futex-2025-07-29' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: selftests/futex: Fix spelling mistake "Succeffuly" -> "Successfully" selftests/futex: Define SYS_futex on 32-bit architectures with 64-bit time_t perf bench futex: Remove support for IMMUTABLE selftests/futex: Remove support for IMMUTABLE futex: Remove support for IMMUTABLE futex: Make futex_private_hash_get() static futex: Use RCU-based per-CPU reference counting instead of rcuref_t selftests/futex: Adapt the private hash test to RCU related changes 
Pull futex updates from Thomas Gleixner:

 - Switch the reference counting to a RCU based per-CPU reference to
   address a performance bottleneck vs the single instance rcuref
   variant

 - Make the futex selftest build on 32-bit architectures which only
   support 64-bit time_t, e.g. RISCV-32

 - Cleanups and improvements in selftests and futex bench

* tag 'locking-futex-2025-07-29' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  selftests/futex: Fix spelling mistake "Succeffuly" -> "Successfully"
  selftests/futex: Define SYS_futex on 32-bit architectures with 64-bit time_t
  perf bench futex: Remove support for IMMUTABLE
  selftests/futex: Remove support for IMMUTABLE
  futex: Remove support for IMMUTABLE
  futex: Make futex_private_hash_get() static
  futex: Use RCU-based per-CPU reference counting instead of rcuref_t
  selftests/futex: Adapt the private hash test to RCU related changes
selftests/futex: Define SYS_futex on 32-bit architectures with 64-bit time_t 2025-07-22T09:18:43+00:00 Cynthia Huang cynthia@andestech.com 2025-07-10T10:36:30+00:00 04850819c65c8242072818655d4341e70ae998b5 The kernel does not provide sys_futex() on 32-bit architectures that do not support 32-bit time representations, such as riscv32. As a result, glibc cannot define SYS_futex, causing compilation failures in tests that rely on this syscall. Define SYS_futex as SYS_futex_time64 in such cases to ensure successful compilation and compatibility. Signed-off-by: Cynthia Huang <cynthia@andestech.com> Signed-off-by: Ben Zong-You Xie <ben717@andestech.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Muhammad Usama Anjum <usama.anjum@collabora.com> Link: https://lore.kernel.org/all/20250710103630.3156130-1-ben717@andestech.com 
The kernel does not provide sys_futex() on 32-bit architectures that do not
support 32-bit time representations, such as riscv32.

As a result, glibc cannot define SYS_futex, causing compilation failures in
tests that rely on this syscall. Define SYS_futex as SYS_futex_time64 in
such cases to ensure successful compilation and compatibility.

Signed-off-by: Cynthia Huang <cynthia@andestech.com>
Signed-off-by: Ben Zong-You Xie <ben717@andestech.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Muhammad Usama Anjum <usama.anjum@collabora.com>
Link: https://lore.kernel.org/all/20250710103630.3156130-1-ben717@andestech.com
selftests/futex: Convert 32-bit timespec to 64-bit version for 32-bit compatibility mode 2025-07-06T09:15:29+00:00 Terry Tritton terry.tritton@linaro.org 2025-07-04T19:02:34+00:00 d0a48dc4df5c986bf8c3caf4d8fc15c480273052 sys_futex_wait() expects a struct __kernel_timespec pointer for the timeout, but the provided struct timespec pointer is of type struct old_timespec32 when compiled for 32-bit architectures, unless they use 64-bit timespecs already. Make it work for all variants by converting the provided timespec value into a local struct __kernel_timespec and provide a pointer to it to the syscall. This is a pointless operation for 64-bit, but this is not a hotpath operation, so keep it simple. This fix is based off [1] Originally-by: Wei Gao <wegao@suse.com> Signed-off-by: Terry Tritton <terry.tritton@linaro.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Link: https://lore.kernel.org/all/20250704190234.14230-1-terry.tritton@linaro.org Link: https://lore.kernel.org/all/20231203235117.29677-1-wegao@suse.com/ [1] 
sys_futex_wait() expects a struct __kernel_timespec pointer for the
timeout, but the provided struct timespec pointer is of type struct
old_timespec32 when compiled for 32-bit architectures, unless they use
64-bit timespecs already.

Make it work for all variants by converting the provided timespec value
into a local struct __kernel_timespec and provide a pointer to it to the
syscall. This is a pointless operation for 64-bit, but this is not a
hotpath operation, so keep it simple.

This fix is based off [1]

Originally-by: Wei Gao <wegao@suse.com>
Signed-off-by: Terry Tritton <terry.tritton@linaro.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20250704190234.14230-1-terry.tritton@linaro.org
Link: https://lore.kernel.org/all/20231203235117.29677-1-wegao@suse.com/ [1]
selftests/futex: Add futex_numa_mpol 2025-05-03T10:02:10+00:00 Sebastian Andrzej Siewior bigeasy@linutronix.de 2025-04-16T16:29:21+00:00 3163369407baf8331a234fe4817e9ea27ba7ea9c Test the basic functionality for the NUMA and MPOL flags: - FUTEX2_NUMA should take the NUMA node which is after the uaddr and use it. - Only update the node if FUTEX_NO_NODE was set by the user - FUTEX2_MPOL should use the memory based on the policy. I attempted to set the node with mbind() and then use this with MPOL but this fails and futex falls back to the default node for the current CPU. Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lore.kernel.org/r/20250416162921.513656-22-bigeasy@linutronix.de 
Test the basic functionality for the NUMA and MPOL flags:
- FUTEX2_NUMA should take the NUMA node which is after the uaddr
  and use it.
- Only update the node if FUTEX_NO_NODE was set by the user
- FUTEX2_MPOL should use the memory based on the policy. I attempted to
  set the node with mbind() and then use this with MPOL but this fails
  and futex falls back to the default node for the current CPU.

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20250416162921.513656-22-bigeasy@linutronix.de
selftests/futex: Build without headers nonsense 2025-05-03T10:02:10+00:00 Peter Zijlstra peterz@infradead.org 2025-05-02T18:57:23+00:00 8b4a5c2497fad653bc54ddb037d38eb5bf835857 Make it build without relying on recent headers. Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> 
Make it build without relying on recent headers.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>