Now that there are no users of smp_mb__before_spinlock() left, remove it entirely. Signed-off-by: Peter Zijlstra (Intel) --- Documentation/memory-barriers.txt | 5 --- Documentation/translations/ko_KR/memory-barriers.txt | 5 --- arch/arm64/include/asm/spinlock.h | 9 ------ arch/powerpc/include/asm/barrier.h | 2 - fs/userfaultfd.c | 25 ++++++++----------- include/linux/spinlock.h | 13 --------- 6 files changed, 13 insertions(+), 46 deletions(-) --- a/Documentation/memory-barriers.txt +++ b/Documentation/memory-barriers.txt @@ -1982,10 +1982,7 @@ In all cases there are variants on "ACQU ACQUIRE operation has completed. Memory operations issued before the ACQUIRE may be completed after - the ACQUIRE operation has completed. An smp_mb__before_spinlock(), - combined with a following ACQUIRE, orders prior stores against - subsequent loads and stores. Note that this is weaker than smp_mb()! - The smp_mb__before_spinlock() primitive is free on many architectures. + the ACQUIRE operation has completed. (2) RELEASE operation implication: --- a/Documentation/translations/ko_KR/memory-barriers.txt +++ b/Documentation/translations/ko_KR/memory-barriers.txt @@ -1956,10 +1956,7 @@ MMIO 쓰기 배리어 뒤에 완료됩니다. ACQUIRE 앞에서 요청된 메모리 오퍼레이션은 ACQUIRE 오퍼레이션이 완료된 후에 - 완료될 수 있습니다. smp_mb__before_spinlock() 뒤에 ACQUIRE 가 실행되는 - 코드 블록은 블록 앞의 스토어를 블록 뒤의 로드와 스토어에 대해 순서 - 맞춥니다. 이건 smp_mb() 보다 완화된 것임을 기억하세요! 많은 아키텍쳐에서 - smp_mb__before_spinlock() 은 사실 아무일도 하지 않습니다. + 완료될 수 있습니다. (2) RELEASE 오퍼레이션의 영향: --- a/arch/arm64/include/asm/spinlock.h +++ b/arch/arm64/include/asm/spinlock.h @@ -358,15 +358,6 @@ static inline int arch_read_trylock(arch #define arch_read_relax(lock) cpu_relax() #define arch_write_relax(lock) cpu_relax() -/* - * Accesses appearing in program order before a spin_lock() operation - * can be reordered with accesses inside the critical section, by virtue - * of arch_spin_lock being constructed using acquire semantics. - * - * In cases where this is problematic (e.g. try_to_wake_up), an - * smp_mb__before_spinlock() can restore the required ordering. - */ -#define smp_mb__before_spinlock() smp_mb() /* See include/linux/spinlock.h */ #define smp_mb__after_spinlock() smp_mb() --- a/arch/powerpc/include/asm/barrier.h +++ b/arch/powerpc/include/asm/barrier.h @@ -74,8 +74,6 @@ do { \ ___p1; \ }) -#define smp_mb__before_spinlock() smp_mb() - #include #endif /* _ASM_POWERPC_BARRIER_H */ --- a/fs/userfaultfd.c +++ b/fs/userfaultfd.c @@ -109,27 +109,24 @@ static int userfaultfd_wake_function(wai goto out; WRITE_ONCE(uwq->waken, true); /* - * The implicit smp_mb__before_spinlock in try_to_wake_up() - * renders uwq->waken visible to other CPUs before the task is - * waken. + * The Program-Order guarantees provided by the scheduler + * ensure uwq->waken is visible before the task is woken. */ ret = wake_up_state(wq->private, mode); - if (ret) + if (ret) { /* * Wake only once, autoremove behavior. * - * After the effect of list_del_init is visible to the - * other CPUs, the waitqueue may disappear from under - * us, see the !list_empty_careful() in - * handle_userfault(). try_to_wake_up() has an - * implicit smp_mb__before_spinlock, and the - * wq->private is read before calling the extern - * function "wake_up_state" (which in turns calls - * try_to_wake_up). While the spin_lock;spin_unlock; - * wouldn't be enough, the smp_mb__before_spinlock is - * enough to avoid an explicit smp_mb() here. + * After the effect of list_del_init is visible to the other + * CPUs, the waitqueue may disappear from under us, see the + * !list_empty_careful() in handle_userfault(). + * + * try_to_wake_up() has an implicit smp_mb(), and the + * wq->private is read before calling the extern function + * "wake_up_state" (which in turns calls try_to_wake_up). */ list_del_init(&wq->entry); + } out: return ret; } --- a/include/linux/spinlock.h +++ b/include/linux/spinlock.h @@ -118,19 +118,6 @@ do { \ #endif /* - * Despite its name it doesn't necessarily has to be a full barrier. - * It should only guarantee that a STORE before the critical section - * can not be reordered with LOADs and STOREs inside this section. - * spin_lock() is the one-way barrier, this LOAD can not escape out - * of the region. So the default implementation simply ensures that - * a STORE can not move into the critical section, smp_wmb() should - * serialize it with another STORE done by spin_lock(). - */ -#ifndef smp_mb__before_spinlock -#define smp_mb__before_spinlock() smp_wmb() -#endif - -/* * This barrier must provide two things: * * - it must guarantee a STORE before the spin_lock() is ordered against a