On Sun, Jan 11, 2026 at 5:34 PM Jonathan Wakely <[email protected]> wrote:
> This defines __platform_wait, __platform_notify, and
> __platform_wait_until for FreeBSD, making use of the _umtx_op syscall.
>
> The Linux versions of those functions only support 32-bit integers, but
> the FreeBSD versions use the syscall for both 32-bit and 64-bit types,
> as the _umtx_op supports both.
>
> libstdc++-v3/ChangeLog:
>
> PR libstdc++/120527
> * include/bits/atomic_wait.h [__FreeBSD__] (__platform_wait_t):
> Define typedef.
> [__FreeBSD__] (__platform_wait_uses_type): Define variable
> template.
> * src/c++20/atomic.cc [__FreeBSD__] (_GLIBCXX_HAVE_PLATFORM_WAIT)
> (__platform_wait, __platform_notify, __platform_wait_until):
> Define.
> (__platform_load): Define.
> ---
>
> v4: Tomasz noticed that __spin_impl assumes the waitable is always a
> __platform_wait_t, but the FreeBSD patch enables the fast path for both
> 32-bit and 64-bit integers, so we need to adjust how __spin_impl loads
> the current value of the waitable. This adds a __platform_load function,
>
which does an atomic load of the right size. For Linux, it's always
> loading from int* but for FreeBSD it might be a 4B or 8B type.
>
I think this should go to patch description, or preferably follow what you
did
in previous patches, and put introduction into __platform_load into
separate commit,
explaining why we need it.
Outside of that LGTM.
>
> Tested x86_64-linux and x86_64-freebsd.
>
> I see a few unexplained FAILs when testing with -m32 on freebsd, but
> they're already present on trunk so aren't caused by this patch. This
> passes all tests with -m64 (and the changes should only affect 64-bit
> more, because of the __SIZEOF_LONG__ == 8 check).
>
> libstdc++-v3/include/bits/atomic_wait.h | 11 ++++
> libstdc++-v3/src/c++20/atomic.cc | 78 ++++++++++++++++++++++++-
> 2 files changed, 87 insertions(+), 2 deletions(-)
>
> diff --git a/libstdc++-v3/include/bits/atomic_wait.h
> b/libstdc++-v3/include/bits/atomic_wait.h
> index eff1be604eb4..e8487390ecb5 100644
> --- a/libstdc++-v3/include/bits/atomic_wait.h
> +++ b/libstdc++-v3/include/bits/atomic_wait.h
> @@ -69,6 +69,17 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
> inline constexpr bool __platform_wait_uses_type
> = __detail::__waitable<_Tp>
> && sizeof(_Tp) == sizeof(int) && alignof(_Tp) >= 4;
> +#elif defined __FreeBSD__ && __SIZEOF_LONG__ == 8
> + namespace __detail
> + {
> + using __platform_wait_t = __UINT64_TYPE__;
> + inline constexpr size_t __platform_wait_alignment = 8;
> + }
> + template<typename _Tp>
> + inline constexpr bool __platform_wait_uses_type
> + = __detail::__waitable<_Tp>
> + && ((sizeof(_Tp) == 4 && alignof(_Tp) >= 4)
> + || (sizeof(_Tp) == 8 && alignof(_Tp) >= 8));
> #else
> // define _GLIBCX_HAVE_PLATFORM_WAIT and implement __platform_wait()
> // and __platform_notify() if there is a more efficient primitive
> supported
> diff --git a/libstdc++-v3/src/c++20/atomic.cc
> b/libstdc++-v3/src/c++20/atomic.cc
> index 3167592786e1..215f1212dd99 100644
> --- a/libstdc++-v3/src/c++20/atomic.cc
> +++ b/libstdc++-v3/src/c++20/atomic.cc
> @@ -27,7 +27,7 @@
> #if __glibcxx_atomic_wait
> #include <atomic>
> #include <bits/atomic_timed_wait.h>
> -#include <cstdint> // uint32_t, uint64_t
> +#include <cstdint> // uint32_t, uint64_t, uintptr_t
> #include <climits> // INT_MAX
> #include <cerrno> // errno, ETIMEDOUT, etc.
> #include <bits/std_mutex.h> // std::mutex, std::__condvar
> @@ -39,6 +39,11 @@
> # include <unistd.h>
> # include <sys/time.h> // timespec
> # define _GLIBCXX_HAVE_PLATFORM_WAIT 1
> +#elif defined __FreeBSD__ && __FreeBSD__ >= 11 && __SIZEOF_LONG__ == 8
> +# include <sys/types.h>
> +# include <sys/umtx.h>
> +# include <sys/time.h>
> +# define _GLIBCXX_HAVE_PLATFORM_WAIT 1
> #endif
>
> #pragma GCC diagnostic ignored "-Wmissing-field-initializers"
> @@ -87,6 +92,13 @@ namespace
> __wait_clock_t::time_point timeout,
> int obj_size) = delete;
>
> + // This is needed even when we don't have __platform_wait
> + [[gnu::always_inline]]
> + inline __wait_value_type
> + __platform_load(const __platform_wait_t* addr, int memory_order,
> + int /* obj_sz */) noexcept
> + { return __atomic_load_n(addr, memory_order); }
> +
> #elif defined _GLIBCXX_HAVE_LINUX_FUTEX
>
> const int futex_private_flag = 128;
> @@ -136,6 +148,68 @@ namespace
> }
> return true;
> }
> +
> + [[gnu::always_inline]]
> + inline __wait_value_type
> + __platform_load(const int* addr, int order, int /* obj_sz */) noexcept
>
I checked that we use int for other __platform functions with futex patch.
> + { return __atomic_load_n(addr, order); }
> +
> +#elif defined __FreeBSD__ && __SIZEOF_LONG__ == 8
> + [[gnu::always_inline]]
> + inline int
> + wait_op(int obj_sz) noexcept
> + { return obj_sz == sizeof(unsigned) ? UMTX_OP_WAIT_UINT : UMTX_OP_WAIT;
> }
> +
> + void
> + __platform_wait(const void* addr, uint64_t val, int obj_sz) noexcept
> + {
> + if (_umtx_op(const_cast<void*>(addr), wait_op(obj_sz), val,
> + nullptr, nullptr))
> + if (errno != EINTR)
> + __throw_system_error(errno);
> + }
> +
> + void
> + __platform_notify(const void* addr, bool all, int /* obj_sz */) noexcept
> + {
> + const int count = all ? INT_MAX : 1;
> + _umtx_op(const_cast<void*>(addr), UMTX_OP_WAKE, count, nullptr,
> nullptr);
> + }
> +
> + // returns true if wait ended before timeout
> + bool
> + __platform_wait_until(const void* addr, uint64_t val,
> + const __wait_clock_t::time_point& atime,
> + int obj_sz) noexcept
> + {
> + struct _umtx_time timeout = {
> + ._timeout = chrono::__to_timeout_timespec(atime),
> + ._flags = UMTX_ABSTIME,
> + ._clockid = CLOCK_MONOTONIC
> + };
> + // _umtx_op hangs if timeout._timeout is {0, 0}
> + if (atime.time_since_epoch() < chrono::nanoseconds(1))
> + return false;
> + constexpr uintptr_t timeout_sz = sizeof(timeout);
> + if (_umtx_op(const_cast<void*>(addr), wait_op(obj_sz), val,
> + (void*)timeout_sz, &timeout))
> + {
> + if (errno == ETIMEDOUT)
> + return false;
> + if (errno != EINTR)
> + __throw_system_error(errno);
> + }
> + return true;
> + }
> +
> + [[gnu::always_inline]]
> + inline __wait_value_type
> + __platform_load(const void* addr, int order, int obj_sz) noexcept
> + {
> + if (obj_sz == sizeof(long))
> + return __atomic_load_n(static_cast<const long*>(addr), order);
> + return __atomic_load_n(static_cast<const unsigned*>(addr), order);
> + }
> #endif // HAVE_PLATFORM_WAIT
>
> // The state used by atomic waiting and notifying functions.
> @@ -259,7 +333,7 @@ namespace
> __wait_value_type wval;
> for (auto i = 0; i < atomic_spin_count; ++i)
> {
> - wval = __atomic_load_n(addr, args._M_order);
> + wval = __platform_load(addr, args._M_order, args._M_obj_size);
>
I was thinking if we should have separate __spin_impl, which would hois the
check
outside of the loop, but this is anyway busy, wait, and the compiler sees
the __platform_load
body. And the __platform_load solution, does not add any if on platform
where waiting only
on single size is supported.
> if (wval != args._M_old)
> return { ._M_val = wval, ._M_has_val = true, ._M_timeout = false
> };
> if (i < atomic_spin_count_relax)
> --
> 2.52.0
>
>
