From 9f61521c7a284e799050cd2adacc9a611bd2b491 Mon Sep 17 00:00:00 2001 From: Nicholas Piggin Date: Mon, 28 Nov 2022 13:11:13 +1000 Subject: powerpc/qspinlock: powerpc qspinlock implementation Add a powerpc specific implementation of queued spinlocks. This is the build framework with a very simple (non-queued) spinlock implementation to begin with. Later changes add queueing, and other features and optimisations one-at-a-time. It is done this way to more easily see how the queued spinlocks are built, and to make performance and correctness bisects more useful. Signed-off-by: Nicholas Piggin [mpe: Drop paravirt.h & processor.h changes to fix 32-bit build] [mpe: Fix 32-bit build of qspinlock.o & disallow GENERIC_LOCKBREAK per Nick] Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/CONLLQB6DCJU.2ZPOS7T6S5GRR@bobo --- arch/powerpc/lib/Makefile | 4 +++- arch/powerpc/lib/qspinlock.c | 17 +++++++++++++++++ 2 files changed, 20 insertions(+), 1 deletion(-) create mode 100644 arch/powerpc/lib/qspinlock.c (limited to 'arch/powerpc/lib') diff --git a/arch/powerpc/lib/Makefile b/arch/powerpc/lib/Makefile index 8560c912186d..4de71cbf6e8e 100644 --- a/arch/powerpc/lib/Makefile +++ b/arch/powerpc/lib/Makefile @@ -52,7 +52,9 @@ obj-$(CONFIG_PPC_BOOK3S_64) += copyuser_power7.o copypage_power7.o \ obj64-y += copypage_64.o copyuser_64.o mem_64.o hweight_64.o \ memcpy_64.o copy_mc_64.o -ifndef CONFIG_PPC_QUEUED_SPINLOCKS +ifdef CONFIG_PPC_QUEUED_SPINLOCKS +obj-$(CONFIG_SMP) += qspinlock.o +else obj64-$(CONFIG_SMP) += locks.o endif diff --git a/arch/powerpc/lib/qspinlock.c b/arch/powerpc/lib/qspinlock.c new file mode 100644 index 000000000000..1c669b5b4607 --- /dev/null +++ b/arch/powerpc/lib/qspinlock.c @@ -0,0 +1,17 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +#include +#include +#include + +void queued_spin_lock_slowpath(struct qspinlock *lock) +{ + while (!queued_spin_trylock(lock)) + cpu_relax(); +} +EXPORT_SYMBOL(queued_spin_lock_slowpath); + +#ifdef CONFIG_PARAVIRT_SPINLOCKS +void pv_spinlocks_init(void) +{ +} +#endif -- cgit v1.2.3 From 84990b169557428c318df87b7836cd15f65b62dc Mon Sep 17 00:00:00 2001 From: Nicholas Piggin Date: Sat, 26 Nov 2022 19:59:16 +1000 Subject: powerpc/qspinlock: add mcs queueing for contended waiters This forms the basis of the qspinlock slow path. Like generic qspinlocks and unlike the vanilla MCS algorithm, the lock owner does not participate in the queue, only waiters. The first waiter spins on the lock word, then when the lock is released it takes ownership and unqueues the next waiter. This is how qspinlocks can be implemented with the spinlock API -- lock owners don't need a node, only waiters do. Signed-off-by: Nicholas Piggin Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20221126095932.1234527-2-npiggin@gmail.com --- arch/powerpc/include/asm/qspinlock.h | 10 +- arch/powerpc/include/asm/qspinlock_types.h | 23 ++++ arch/powerpc/lib/qspinlock.c | 187 ++++++++++++++++++++++++++++- 3 files changed, 214 insertions(+), 6 deletions(-) (limited to 'arch/powerpc/lib') diff --git a/arch/powerpc/include/asm/qspinlock.h b/arch/powerpc/include/asm/qspinlock.h index 5e6257313557..6946dba5d087 100644 --- a/arch/powerpc/include/asm/qspinlock.h +++ b/arch/powerpc/include/asm/qspinlock.h @@ -19,12 +19,12 @@ static __always_inline int queued_spin_value_unlocked(struct qspinlock lock) static __always_inline int queued_spin_is_contended(struct qspinlock *lock) { - return 0; + return !!(atomic_read(&lock->val) & _Q_TAIL_CPU_MASK); } static __always_inline int queued_spin_trylock(struct qspinlock *lock) { - return atomic_cmpxchg_acquire(&lock->val, 0, 1) == 0; + return atomic_cmpxchg_acquire(&lock->val, 0, _Q_LOCKED_VAL) == 0; } void queued_spin_lock_slowpath(struct qspinlock *lock); @@ -37,7 +37,11 @@ static __always_inline void queued_spin_lock(struct qspinlock *lock) static inline void queued_spin_unlock(struct qspinlock *lock) { - atomic_set_release(&lock->val, 0); + for (;;) { + int val = atomic_read(&lock->val); + if (atomic_cmpxchg_release(&lock->val, val, val & ~_Q_LOCKED_VAL) == val) + return; + } } #define arch_spin_is_locked(l) queued_spin_is_locked(l) diff --git a/arch/powerpc/include/asm/qspinlock_types.h b/arch/powerpc/include/asm/qspinlock_types.h index 59606bc0c774..20a36dfb14e2 100644 --- a/arch/powerpc/include/asm/qspinlock_types.h +++ b/arch/powerpc/include/asm/qspinlock_types.h @@ -10,4 +10,27 @@ typedef struct qspinlock { #define __ARCH_SPIN_LOCK_UNLOCKED { .val = ATOMIC_INIT(0) } +/* + * Bitfields in the lock word: + * + * 0: locked bit + * 1-16: unused bits + * 17-31: tail cpu (+1) + */ +#define _Q_SET_MASK(type) (((1U << _Q_ ## type ## _BITS) - 1)\ + << _Q_ ## type ## _OFFSET) +/* 0x00000001 */ +#define _Q_LOCKED_OFFSET 0 +#define _Q_LOCKED_BITS 1 +#define _Q_LOCKED_VAL (1U << _Q_LOCKED_OFFSET) + +/* 0xfffe0000 */ +#define _Q_TAIL_CPU_OFFSET 17 +#define _Q_TAIL_CPU_BITS 15 +#define _Q_TAIL_CPU_MASK _Q_SET_MASK(TAIL_CPU) + +#if CONFIG_NR_CPUS >= (1U << _Q_TAIL_CPU_BITS) +#error "qspinlock does not support such large CONFIG_NR_CPUS" +#endif + #endif /* _ASM_POWERPC_QSPINLOCK_TYPES_H */ diff --git a/arch/powerpc/lib/qspinlock.c b/arch/powerpc/lib/qspinlock.c index 1c669b5b4607..86504628501e 100644 --- a/arch/powerpc/lib/qspinlock.c +++ b/arch/powerpc/lib/qspinlock.c @@ -1,12 +1,193 @@ // SPDX-License-Identifier: GPL-2.0-or-later +#include +#include +#include #include -#include +#include +#include #include -void queued_spin_lock_slowpath(struct qspinlock *lock) +#define MAX_NODES 4 + +struct qnode { + struct qnode *next; + struct qspinlock *lock; + u8 locked; /* 1 if lock acquired */ +}; + +struct qnodes { + int count; + struct qnode nodes[MAX_NODES]; +}; + +static DEFINE_PER_CPU_ALIGNED(struct qnodes, qnodes); + +static inline int encode_tail_cpu(int cpu) +{ + return (cpu + 1) << _Q_TAIL_CPU_OFFSET; +} + +static inline int decode_tail_cpu(int val) +{ + return (val >> _Q_TAIL_CPU_OFFSET) - 1; +} + +/* + * Try to acquire the lock if it was not already locked. If the tail matches + * mytail then clear it, otherwise leave it unchnaged. Return previous value. + * + * This is used by the head of the queue to acquire the lock and clean up + * its tail if it was the last one queued. + */ +static __always_inline int set_locked_clean_tail(struct qspinlock *lock, int tail) +{ + int val = atomic_read(&lock->val); + + BUG_ON(val & _Q_LOCKED_VAL); + + /* If we're the last queued, must clean up the tail. */ + if ((val & _Q_TAIL_CPU_MASK) == tail) { + if (atomic_cmpxchg_acquire(&lock->val, val, _Q_LOCKED_VAL) == val) + return val; + /* Another waiter must have enqueued */ + val = atomic_read(&lock->val); + BUG_ON(val & _Q_LOCKED_VAL); + } + + /* We must be the owner, just set the lock bit and acquire */ + atomic_or(_Q_LOCKED_VAL, &lock->val); + __atomic_acquire_fence(); + + return val; +} + +/* + * Publish our tail, replacing previous tail. Return previous value. + * + * This provides a release barrier for publishing node, this pairs with the + * acquire barrier in get_tail_qnode() when the next CPU finds this tail + * value. + */ +static __always_inline int publish_tail_cpu(struct qspinlock *lock, int tail) +{ + for (;;) { + int val = atomic_read(&lock->val); + int newval = (val & ~_Q_TAIL_CPU_MASK) | tail; + int old; + + old = atomic_cmpxchg_release(&lock->val, val, newval); + if (old == val) + return old; + } +} + +static struct qnode *get_tail_qnode(struct qspinlock *lock, int val) +{ + int cpu = decode_tail_cpu(val); + struct qnodes *qnodesp = per_cpu_ptr(&qnodes, cpu); + int idx; + + /* + * After publishing the new tail and finding a previous tail in the + * previous val (which is the control dependency), this barrier + * orders the release barrier in publish_tail_cpu performed by the + * last CPU, with subsequently looking at its qnode structures + * after the barrier. + */ + smp_acquire__after_ctrl_dep(); + + for (idx = 0; idx < MAX_NODES; idx++) { + struct qnode *qnode = &qnodesp->nodes[idx]; + if (qnode->lock == lock) + return qnode; + } + + BUG(); +} + +static inline void queued_spin_lock_mcs_queue(struct qspinlock *lock) { - while (!queued_spin_trylock(lock)) + struct qnodes *qnodesp; + struct qnode *next, *node; + int val, old, tail; + int idx; + + BUILD_BUG_ON(CONFIG_NR_CPUS >= (1U << _Q_TAIL_CPU_BITS)); + + qnodesp = this_cpu_ptr(&qnodes); + if (unlikely(qnodesp->count >= MAX_NODES)) { + while (!queued_spin_trylock(lock)) + cpu_relax(); + return; + } + + idx = qnodesp->count++; + /* + * Ensure that we increment the head node->count before initialising + * the actual node. If the compiler is kind enough to reorder these + * stores, then an IRQ could overwrite our assignments. + */ + barrier(); + node = &qnodesp->nodes[idx]; + node->next = NULL; + node->lock = lock; + node->locked = 0; + + tail = encode_tail_cpu(smp_processor_id()); + + old = publish_tail_cpu(lock, tail); + + /* + * If there was a previous node; link it and wait until reaching the + * head of the waitqueue. + */ + if (old & _Q_TAIL_CPU_MASK) { + struct qnode *prev = get_tail_qnode(lock, old); + + /* Link @node into the waitqueue. */ + WRITE_ONCE(prev->next, node); + + /* Wait for mcs node lock to be released */ + while (!node->locked) + cpu_relax(); + + smp_rmb(); /* acquire barrier for the mcs lock */ + } + + /* We're at the head of the waitqueue, wait for the lock. */ + for (;;) { + val = atomic_read(&lock->val); + if (!(val & _Q_LOCKED_VAL)) + break; + + cpu_relax(); + } + + /* If we're the last queued, must clean up the tail. */ + old = set_locked_clean_tail(lock, tail); + if ((old & _Q_TAIL_CPU_MASK) == tail) + goto release; /* Another waiter must have enqueued */ + + /* There is a next, must wait for node->next != NULL (MCS protocol) */ + while (!(next = READ_ONCE(node->next))) cpu_relax(); + + /* + * Unlock the next mcs waiter node. Release barrier is not required + * here because the acquirer is only accessing the lock word, and + * the acquire barrier we took the lock with orders that update vs + * this store to locked. The corresponding barrier is the smp_rmb() + * acquire barrier for mcs lock, above. + */ + WRITE_ONCE(next->locked, 1); + +release: + qnodesp->count--; /* release the node */ +} + +void queued_spin_lock_slowpath(struct qspinlock *lock) +{ + queued_spin_lock_mcs_queue(lock); } EXPORT_SYMBOL(queued_spin_lock_slowpath); -- cgit v1.2.3 From b3a73b7db2b6cb3b2e5bfda5518a0e92230ef673 Mon Sep 17 00:00:00 2001 From: Nicholas Piggin Date: Sat, 26 Nov 2022 19:59:18 +1000 Subject: powerpc/qspinlock: convert atomic operations to assembly This uses more optimal ll/sc style access patterns (rather than cmpxchg), and also sets the EH=1 lock hint on those operations which acquire ownership of the lock. Signed-off-by: Nicholas Piggin Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20221126095932.1234527-4-npiggin@gmail.com --- arch/powerpc/include/asm/qspinlock.h | 24 +++++++-- arch/powerpc/include/asm/qspinlock_types.h | 4 +- arch/powerpc/lib/qspinlock.c | 82 +++++++++++++++++------------- 3 files changed, 68 insertions(+), 42 deletions(-) (limited to 'arch/powerpc/lib') diff --git a/arch/powerpc/include/asm/qspinlock.h b/arch/powerpc/include/asm/qspinlock.h index 713f6629f6fb..c16e1f0674b5 100644 --- a/arch/powerpc/include/asm/qspinlock.h +++ b/arch/powerpc/include/asm/qspinlock.h @@ -2,29 +2,43 @@ #ifndef _ASM_POWERPC_QSPINLOCK_H #define _ASM_POWERPC_QSPINLOCK_H -#include #include #include #include static __always_inline int queued_spin_is_locked(struct qspinlock *lock) { - return atomic_read(&lock->val); + return READ_ONCE(lock->val); } static __always_inline int queued_spin_value_unlocked(struct qspinlock lock) { - return !atomic_read(&lock.val); + return !lock.val; } static __always_inline int queued_spin_is_contended(struct qspinlock *lock) { - return !!(atomic_read(&lock->val) & _Q_TAIL_CPU_MASK); + return !!(READ_ONCE(lock->val) & _Q_TAIL_CPU_MASK); } static __always_inline int queued_spin_trylock(struct qspinlock *lock) { - return atomic_cmpxchg_acquire(&lock->val, 0, _Q_LOCKED_VAL) == 0; + u32 prev; + + asm volatile( +"1: lwarx %0,0,%1,%3 # queued_spin_trylock \n" +" cmpwi 0,%0,0 \n" +" bne- 2f \n" +" stwcx. %2,0,%1 \n" +" bne- 1b \n" +"\t" PPC_ACQUIRE_BARRIER " \n" +"2: \n" + : "=&r" (prev) + : "r" (&lock->val), "r" (_Q_LOCKED_VAL), + "i" (IS_ENABLED(CONFIG_PPC64)) + : "cr0", "memory"); + + return likely(prev == 0); } void queued_spin_lock_slowpath(struct qspinlock *lock); diff --git a/arch/powerpc/include/asm/qspinlock_types.h b/arch/powerpc/include/asm/qspinlock_types.h index fe87181c59e5..b9a5a52fa670 100644 --- a/arch/powerpc/include/asm/qspinlock_types.h +++ b/arch/powerpc/include/asm/qspinlock_types.h @@ -7,7 +7,7 @@ typedef struct qspinlock { union { - atomic_t val; + u32 val; #ifdef __LITTLE_ENDIAN struct { @@ -23,7 +23,7 @@ typedef struct qspinlock { }; } arch_spinlock_t; -#define __ARCH_SPIN_LOCK_UNLOCKED { { .val = ATOMIC_INIT(0) } } +#define __ARCH_SPIN_LOCK_UNLOCKED { { .val = 0 } } /* * Bitfields in the lock word: diff --git a/arch/powerpc/lib/qspinlock.c b/arch/powerpc/lib/qspinlock.c index 86504628501e..645d9affacfd 100644 --- a/arch/powerpc/lib/qspinlock.c +++ b/arch/powerpc/lib/qspinlock.c @@ -1,5 +1,4 @@ // SPDX-License-Identifier: GPL-2.0-or-later -#include #include #include #include @@ -22,12 +21,12 @@ struct qnodes { static DEFINE_PER_CPU_ALIGNED(struct qnodes, qnodes); -static inline int encode_tail_cpu(int cpu) +static inline u32 encode_tail_cpu(int cpu) { return (cpu + 1) << _Q_TAIL_CPU_OFFSET; } -static inline int decode_tail_cpu(int val) +static inline int decode_tail_cpu(u32 val) { return (val >> _Q_TAIL_CPU_OFFSET) - 1; } @@ -39,26 +38,34 @@ static inline int decode_tail_cpu(int val) * This is used by the head of the queue to acquire the lock and clean up * its tail if it was the last one queued. */ -static __always_inline int set_locked_clean_tail(struct qspinlock *lock, int tail) +static __always_inline u32 set_locked_clean_tail(struct qspinlock *lock, u32 tail) { - int val = atomic_read(&lock->val); - - BUG_ON(val & _Q_LOCKED_VAL); - - /* If we're the last queued, must clean up the tail. */ - if ((val & _Q_TAIL_CPU_MASK) == tail) { - if (atomic_cmpxchg_acquire(&lock->val, val, _Q_LOCKED_VAL) == val) - return val; - /* Another waiter must have enqueued */ - val = atomic_read(&lock->val); - BUG_ON(val & _Q_LOCKED_VAL); - } - - /* We must be the owner, just set the lock bit and acquire */ - atomic_or(_Q_LOCKED_VAL, &lock->val); - __atomic_acquire_fence(); - - return val; + u32 newval = _Q_LOCKED_VAL; + u32 prev, tmp; + + asm volatile( +"1: lwarx %0,0,%2,%6 # set_locked_clean_tail \n" + /* Test whether the lock tail == tail */ +" and %1,%0,%5 \n" +" cmpw 0,%1,%3 \n" + /* Merge the new locked value */ +" or %1,%1,%4 \n" +" bne 2f \n" + /* If the lock tail matched, then clear it, otherwise leave it. */ +" andc %1,%1,%5 \n" +"2: stwcx. %1,0,%2 \n" +" bne- 1b \n" +"\t" PPC_ACQUIRE_BARRIER " \n" +"3: \n" + : "=&r" (prev), "=&r" (tmp) + : "r" (&lock->val), "r"(tail), "r" (newval), + "r" (_Q_TAIL_CPU_MASK), + "i" (IS_ENABLED(CONFIG_PPC64)) + : "cr0", "memory"); + + BUG_ON(prev & _Q_LOCKED_VAL); + + return prev; } /* @@ -68,20 +75,25 @@ static __always_inline int set_locked_clean_tail(struct qspinlock *lock, int tai * acquire barrier in get_tail_qnode() when the next CPU finds this tail * value. */ -static __always_inline int publish_tail_cpu(struct qspinlock *lock, int tail) +static __always_inline u32 publish_tail_cpu(struct qspinlock *lock, u32 tail) { - for (;;) { - int val = atomic_read(&lock->val); - int newval = (val & ~_Q_TAIL_CPU_MASK) | tail; - int old; - - old = atomic_cmpxchg_release(&lock->val, val, newval); - if (old == val) - return old; - } + u32 prev, tmp; + + asm volatile( +"\t" PPC_RELEASE_BARRIER " \n" +"1: lwarx %0,0,%2 # publish_tail_cpu \n" +" andc %1,%0,%4 \n" +" or %1,%1,%3 \n" +" stwcx. %1,0,%2 \n" +" bne- 1b \n" + : "=&r" (prev), "=&r"(tmp) + : "r" (&lock->val), "r" (tail), "r"(_Q_TAIL_CPU_MASK) + : "cr0", "memory"); + + return prev; } -static struct qnode *get_tail_qnode(struct qspinlock *lock, int val) +static struct qnode *get_tail_qnode(struct qspinlock *lock, u32 val) { int cpu = decode_tail_cpu(val); struct qnodes *qnodesp = per_cpu_ptr(&qnodes, cpu); @@ -109,7 +121,7 @@ static inline void queued_spin_lock_mcs_queue(struct qspinlock *lock) { struct qnodes *qnodesp; struct qnode *next, *node; - int val, old, tail; + u32 val, old, tail; int idx; BUILD_BUG_ON(CONFIG_NR_CPUS >= (1U << _Q_TAIL_CPU_BITS)); @@ -156,7 +168,7 @@ static inline void queued_spin_lock_mcs_queue(struct qspinlock *lock) /* We're at the head of the waitqueue, wait for the lock. */ for (;;) { - val = atomic_read(&lock->val); + val = READ_ONCE(lock->val); if (!(val & _Q_LOCKED_VAL)) break; -- cgit v1.2.3 From 6aa42f883c438ea132a28801bef3f86f3883d14c Mon Sep 17 00:00:00 2001 From: Nicholas Piggin Date: Sat, 26 Nov 2022 19:59:19 +1000 Subject: powerpc/qspinlock: allow new waiters to steal the lock before queueing Allow new waiters to "steal" the lock before queueing. That is, to acquire it while other CPUs have queued. This particularly helps paravirt performance when physical CPUs are oversubscribed, by keeping the lock from becoming a strict FIFO and vCPU preemption causing queue train wrecks. The new __queued_spin_trylock_steal() function is put in qspinlock.h to save having to move it, because it will be used there by a later change. Signed-off-by: Nicholas Piggin Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20221126095932.1234527-5-npiggin@gmail.com --- arch/powerpc/include/asm/qspinlock.h | 23 ++++++++ arch/powerpc/lib/qspinlock.c | 110 ++++++++++++++++++++++++++++++++--- 2 files changed, 124 insertions(+), 9 deletions(-) (limited to 'arch/powerpc/lib') diff --git a/arch/powerpc/include/asm/qspinlock.h b/arch/powerpc/include/asm/qspinlock.h index c16e1f0674b5..cebd2c89c08d 100644 --- a/arch/powerpc/include/asm/qspinlock.h +++ b/arch/powerpc/include/asm/qspinlock.h @@ -41,6 +41,29 @@ static __always_inline int queued_spin_trylock(struct qspinlock *lock) return likely(prev == 0); } +static __always_inline int __queued_spin_trylock_steal(struct qspinlock *lock) +{ + u32 prev, tmp; + + /* Trylock may get ahead of queued nodes if it finds unlocked */ + asm volatile( +"1: lwarx %0,0,%2,%5 # __queued_spin_trylock_steal \n" +" andc. %1,%0,%4 \n" +" bne- 2f \n" +" and %1,%0,%4 \n" +" or %1,%1,%3 \n" +" stwcx. %1,0,%2 \n" +" bne- 1b \n" +"\t" PPC_ACQUIRE_BARRIER " \n" +"2: \n" + : "=&r" (prev), "=&r" (tmp) + : "r" (&lock->val), "r" (_Q_LOCKED_VAL), "r" (_Q_TAIL_CPU_MASK), + "i" (IS_ENABLED(CONFIG_PPC64)) + : "cr0", "memory"); + + return likely(!(prev & ~_Q_TAIL_CPU_MASK)); +} + void queued_spin_lock_slowpath(struct qspinlock *lock); static __always_inline void queued_spin_lock(struct qspinlock *lock) diff --git a/arch/powerpc/lib/qspinlock.c b/arch/powerpc/lib/qspinlock.c index 645d9affacfd..6ffd3261064c 100644 --- a/arch/powerpc/lib/qspinlock.c +++ b/arch/powerpc/lib/qspinlock.c @@ -19,8 +19,17 @@ struct qnodes { struct qnode nodes[MAX_NODES]; }; +/* Tuning parameters */ +static int steal_spins __read_mostly = (1 << 5); +static bool maybe_stealers __read_mostly = true; + static DEFINE_PER_CPU_ALIGNED(struct qnodes, qnodes); +static __always_inline int get_steal_spins(void) +{ + return steal_spins; +} + static inline u32 encode_tail_cpu(int cpu) { return (cpu + 1) << _Q_TAIL_CPU_OFFSET; @@ -38,33 +47,35 @@ static inline int decode_tail_cpu(u32 val) * This is used by the head of the queue to acquire the lock and clean up * its tail if it was the last one queued. */ -static __always_inline u32 set_locked_clean_tail(struct qspinlock *lock, u32 tail) +static __always_inline u32 trylock_clean_tail(struct qspinlock *lock, u32 tail) { u32 newval = _Q_LOCKED_VAL; u32 prev, tmp; asm volatile( -"1: lwarx %0,0,%2,%6 # set_locked_clean_tail \n" - /* Test whether the lock tail == tail */ -" and %1,%0,%5 \n" +"1: lwarx %0,0,%2,%7 # trylock_clean_tail \n" + /* This test is necessary if there could be stealers */ +" andi. %1,%0,%5 \n" +" bne 3f \n" + /* Test whether the lock tail == mytail */ +" and %1,%0,%6 \n" " cmpw 0,%1,%3 \n" /* Merge the new locked value */ " or %1,%1,%4 \n" " bne 2f \n" /* If the lock tail matched, then clear it, otherwise leave it. */ -" andc %1,%1,%5 \n" +" andc %1,%1,%6 \n" "2: stwcx. %1,0,%2 \n" " bne- 1b \n" "\t" PPC_ACQUIRE_BARRIER " \n" "3: \n" : "=&r" (prev), "=&r" (tmp) : "r" (&lock->val), "r"(tail), "r" (newval), + "i" (_Q_LOCKED_VAL), "r" (_Q_TAIL_CPU_MASK), "i" (IS_ENABLED(CONFIG_PPC64)) : "cr0", "memory"); - BUG_ON(prev & _Q_LOCKED_VAL); - return prev; } @@ -117,6 +128,30 @@ static struct qnode *get_tail_qnode(struct qspinlock *lock, u32 val) BUG(); } +static inline bool try_to_steal_lock(struct qspinlock *lock) +{ + int iters = 0; + + if (!steal_spins) + return false; + + /* Attempt to steal the lock */ + do { + u32 val = READ_ONCE(lock->val); + + if (unlikely(!(val & _Q_LOCKED_VAL))) { + if (__queued_spin_trylock_steal(lock)) + return true; + } else { + cpu_relax(); + } + + iters++; + } while (iters < get_steal_spins()); + + return false; +} + static inline void queued_spin_lock_mcs_queue(struct qspinlock *lock) { struct qnodes *qnodesp; @@ -166,6 +201,7 @@ static inline void queued_spin_lock_mcs_queue(struct qspinlock *lock) smp_rmb(); /* acquire barrier for the mcs lock */ } +again: /* We're at the head of the waitqueue, wait for the lock. */ for (;;) { val = READ_ONCE(lock->val); @@ -176,9 +212,14 @@ static inline void queued_spin_lock_mcs_queue(struct qspinlock *lock) } /* If we're the last queued, must clean up the tail. */ - old = set_locked_clean_tail(lock, tail); + old = trylock_clean_tail(lock, tail); + if (unlikely(old & _Q_LOCKED_VAL)) { + BUG_ON(!maybe_stealers); + goto again; /* Can only be true if maybe_stealers. */ + } + if ((old & _Q_TAIL_CPU_MASK) == tail) - goto release; /* Another waiter must have enqueued */ + goto release; /* We were the tail, no next. */ /* There is a next, must wait for node->next != NULL (MCS protocol) */ while (!(next = READ_ONCE(node->next))) @@ -199,6 +240,9 @@ release: void queued_spin_lock_slowpath(struct qspinlock *lock) { + if (try_to_steal_lock(lock)) + return; + queued_spin_lock_mcs_queue(lock); } EXPORT_SYMBOL(queued_spin_lock_slowpath); @@ -208,3 +252,51 @@ void pv_spinlocks_init(void) { } #endif + +#include +static int steal_spins_set(void *data, u64 val) +{ + static DEFINE_MUTEX(lock); + + /* + * The lock slow path has a !maybe_stealers case that can assume + * the head of queue will not see concurrent waiters. That waiter + * is unsafe in the presence of stealers, so must keep them away + * from one another. + */ + + mutex_lock(&lock); + if (val && !steal_spins) { + maybe_stealers = true; + /* wait for queue head waiter to go away */ + synchronize_rcu(); + steal_spins = val; + } else if (!val && steal_spins) { + steal_spins = val; + /* wait for all possible stealers to go away */ + synchronize_rcu(); + maybe_stealers = false; + } else { + steal_spins = val; + } + mutex_unlock(&lock); + + return 0; +} + +static int steal_spins_get(void *data, u64 *val) +{ + *val = steal_spins; + + return 0; +} + +DEFINE_SIMPLE_ATTRIBUTE(fops_steal_spins, steal_spins_get, steal_spins_set, "%llu\n"); + +static __init int spinlock_debugfs_init(void) +{ + debugfs_create_file("qspl_steal_spins", 0600, arch_debugfs_dir, NULL, &fops_steal_spins); + + return 0; +} +device_initcall(spinlock_debugfs_init); -- cgit v1.2.3 From 0944534ef4d5cf39c8133575524be0be3337dd62 Mon Sep 17 00:00:00 2001 From: Nicholas Piggin Date: Sat, 26 Nov 2022 19:59:20 +1000 Subject: powerpc/qspinlock: theft prevention to control latency Give the queue head the ability to stop stealers. After a number of spins without successfully acquiring the lock, the queue head sets this, which halts stealing and will assure it is the next owner. Signed-off-by: Nicholas Piggin Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20221126095932.1234527-6-npiggin@gmail.com --- arch/powerpc/include/asm/qspinlock_types.h | 8 ++++- arch/powerpc/lib/qspinlock.c | 53 ++++++++++++++++++++++++++++++ 2 files changed, 60 insertions(+), 1 deletion(-) (limited to 'arch/powerpc/lib') diff --git a/arch/powerpc/include/asm/qspinlock_types.h b/arch/powerpc/include/asm/qspinlock_types.h index b9a5a52fa670..1911a8a16237 100644 --- a/arch/powerpc/include/asm/qspinlock_types.h +++ b/arch/powerpc/include/asm/qspinlock_types.h @@ -29,7 +29,8 @@ typedef struct qspinlock { * Bitfields in the lock word: * * 0: locked bit - * 1-16: unused bits + * 1-15: unused bits + * 16: must queue bit * 17-31: tail cpu (+1) */ #define _Q_SET_MASK(type) (((1U << _Q_ ## type ## _BITS) - 1)\ @@ -39,6 +40,11 @@ typedef struct qspinlock { #define _Q_LOCKED_BITS 1 #define _Q_LOCKED_VAL (1U << _Q_LOCKED_OFFSET) +/* 0x00010000 */ +#define _Q_MUST_Q_OFFSET 16 +#define _Q_MUST_Q_BITS 1 +#define _Q_MUST_Q_VAL (1U << _Q_MUST_Q_OFFSET) + /* 0xfffe0000 */ #define _Q_TAIL_CPU_OFFSET 17 #define _Q_TAIL_CPU_BITS 15 diff --git a/arch/powerpc/lib/qspinlock.c b/arch/powerpc/lib/qspinlock.c index 6ffd3261064c..9cd442d46b9f 100644 --- a/arch/powerpc/lib/qspinlock.c +++ b/arch/powerpc/lib/qspinlock.c @@ -22,6 +22,7 @@ struct qnodes { /* Tuning parameters */ static int steal_spins __read_mostly = (1 << 5); static bool maybe_stealers __read_mostly = true; +static int head_spins __read_mostly = (1 << 8); static DEFINE_PER_CPU_ALIGNED(struct qnodes, qnodes); @@ -30,6 +31,11 @@ static __always_inline int get_steal_spins(void) return steal_spins; } +static __always_inline int get_head_spins(void) +{ + return head_spins; +} + static inline u32 encode_tail_cpu(int cpu) { return (cpu + 1) << _Q_TAIL_CPU_OFFSET; @@ -104,6 +110,22 @@ static __always_inline u32 publish_tail_cpu(struct qspinlock *lock, u32 tail) return prev; } +static __always_inline u32 set_mustq(struct qspinlock *lock) +{ + u32 prev; + + asm volatile( +"1: lwarx %0,0,%1 # set_mustq \n" +" or %0,%0,%2 \n" +" stwcx. %0,0,%1 \n" +" bne- 1b \n" + : "=&r" (prev) + : "r" (&lock->val), "r" (_Q_MUST_Q_VAL) + : "cr0", "memory"); + + return prev; +} + static struct qnode *get_tail_qnode(struct qspinlock *lock, u32 val) { int cpu = decode_tail_cpu(val); @@ -139,6 +161,9 @@ static inline bool try_to_steal_lock(struct qspinlock *lock) do { u32 val = READ_ONCE(lock->val); + if (val & _Q_MUST_Q_VAL) + break; + if (unlikely(!(val & _Q_LOCKED_VAL))) { if (__queued_spin_trylock_steal(lock)) return true; @@ -157,7 +182,9 @@ static inline void queued_spin_lock_mcs_queue(struct qspinlock *lock) struct qnodes *qnodesp; struct qnode *next, *node; u32 val, old, tail; + bool mustq = false; int idx; + int iters = 0; BUILD_BUG_ON(CONFIG_NR_CPUS >= (1U << _Q_TAIL_CPU_BITS)); @@ -209,6 +236,15 @@ again: break; cpu_relax(); + if (!maybe_stealers) + continue; + iters++; + + if (!mustq && iters >= get_head_spins()) { + mustq = true; + set_mustq(lock); + val |= _Q_MUST_Q_VAL; + } } /* If we're the last queued, must clean up the tail. */ @@ -293,9 +329,26 @@ static int steal_spins_get(void *data, u64 *val) DEFINE_SIMPLE_ATTRIBUTE(fops_steal_spins, steal_spins_get, steal_spins_set, "%llu\n"); +static int head_spins_set(void *data, u64 val) +{ + head_spins = val; + + return 0; +} + +static int head_spins_get(void *data, u64 *val) +{ + *val = head_spins; + + return 0; +} + +DEFINE_SIMPLE_ATTRIBUTE(fops_head_spins, head_spins_get, head_spins_set, "%llu\n"); + static __init int spinlock_debugfs_init(void) { debugfs_create_file("qspl_steal_spins", 0600, arch_debugfs_dir, NULL, &fops_steal_spins); + debugfs_create_file("qspl_head_spins", 0600, arch_debugfs_dir, NULL, &fops_head_spins); return 0; } -- cgit v1.2.3 From e1a31e7fd7130628cfd229253da2b4630e7a809c Mon Sep 17 00:00:00 2001 From: Nicholas Piggin Date: Sat, 26 Nov 2022 19:59:21 +1000 Subject: powerpc/qspinlock: store owner CPU in lock word Store the owner CPU number in the lock word so it may be yielded to, as powerpc's paravirtualised simple spinlocks do. Signed-off-by: Nicholas Piggin Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20221126095932.1234527-7-npiggin@gmail.com --- arch/powerpc/include/asm/qspinlock.h | 12 ++++++++++-- arch/powerpc/include/asm/qspinlock_types.h | 12 +++++++++++- arch/powerpc/lib/qspinlock.c | 2 +- 3 files changed, 22 insertions(+), 4 deletions(-) (limited to 'arch/powerpc/lib') diff --git a/arch/powerpc/include/asm/qspinlock.h b/arch/powerpc/include/asm/qspinlock.h index cebd2c89c08d..9572a2ef974d 100644 --- a/arch/powerpc/include/asm/qspinlock.h +++ b/arch/powerpc/include/asm/qspinlock.h @@ -21,8 +21,15 @@ static __always_inline int queued_spin_is_contended(struct qspinlock *lock) return !!(READ_ONCE(lock->val) & _Q_TAIL_CPU_MASK); } +static __always_inline u32 queued_spin_encode_locked_val(void) +{ + /* XXX: make this use lock value in paca like simple spinlocks? */ + return _Q_LOCKED_VAL | (smp_processor_id() << _Q_OWNER_CPU_OFFSET); +} + static __always_inline int queued_spin_trylock(struct qspinlock *lock) { + u32 new = queued_spin_encode_locked_val(); u32 prev; asm volatile( @@ -34,7 +41,7 @@ static __always_inline int queued_spin_trylock(struct qspinlock *lock) "\t" PPC_ACQUIRE_BARRIER " \n" "2: \n" : "=&r" (prev) - : "r" (&lock->val), "r" (_Q_LOCKED_VAL), + : "r" (&lock->val), "r" (new), "i" (IS_ENABLED(CONFIG_PPC64)) : "cr0", "memory"); @@ -43,6 +50,7 @@ static __always_inline int queued_spin_trylock(struct qspinlock *lock) static __always_inline int __queued_spin_trylock_steal(struct qspinlock *lock) { + u32 new = queued_spin_encode_locked_val(); u32 prev, tmp; /* Trylock may get ahead of queued nodes if it finds unlocked */ @@ -57,7 +65,7 @@ static __always_inline int __queued_spin_trylock_steal(struct qspinlock *lock) "\t" PPC_ACQUIRE_BARRIER " \n" "2: \n" : "=&r" (prev), "=&r" (tmp) - : "r" (&lock->val), "r" (_Q_LOCKED_VAL), "r" (_Q_TAIL_CPU_MASK), + : "r" (&lock->val), "r" (new), "r" (_Q_TAIL_CPU_MASK), "i" (IS_ENABLED(CONFIG_PPC64)) : "cr0", "memory"); diff --git a/arch/powerpc/include/asm/qspinlock_types.h b/arch/powerpc/include/asm/qspinlock_types.h index 1911a8a16237..adfeed4aa495 100644 --- a/arch/powerpc/include/asm/qspinlock_types.h +++ b/arch/powerpc/include/asm/qspinlock_types.h @@ -29,7 +29,8 @@ typedef struct qspinlock { * Bitfields in the lock word: * * 0: locked bit - * 1-15: unused bits + * 1-14: lock holder cpu + * 15: unused bit * 16: must queue bit * 17-31: tail cpu (+1) */ @@ -40,6 +41,15 @@ typedef struct qspinlock { #define _Q_LOCKED_BITS 1 #define _Q_LOCKED_VAL (1U << _Q_LOCKED_OFFSET) +/* 0x00007ffe */ +#define _Q_OWNER_CPU_OFFSET 1 +#define _Q_OWNER_CPU_BITS 14 +#define _Q_OWNER_CPU_MASK _Q_SET_MASK(OWNER_CPU) + +#if CONFIG_NR_CPUS > (1U << _Q_OWNER_CPU_BITS) +#error "qspinlock does not support such large CONFIG_NR_CPUS" +#endif + /* 0x00010000 */ #define _Q_MUST_Q_OFFSET 16 #define _Q_MUST_Q_BITS 1 diff --git a/arch/powerpc/lib/qspinlock.c b/arch/powerpc/lib/qspinlock.c index 9cd442d46b9f..4d74db0e565f 100644 --- a/arch/powerpc/lib/qspinlock.c +++ b/arch/powerpc/lib/qspinlock.c @@ -55,7 +55,7 @@ static inline int decode_tail_cpu(u32 val) */ static __always_inline u32 trylock_clean_tail(struct qspinlock *lock, u32 tail) { - u32 newval = _Q_LOCKED_VAL; + u32 newval = queued_spin_encode_locked_val(); u32 prev, tmp; asm volatile( -- cgit v1.2.3 From 085f03311bcede99550e08a1f7cad41bf758b460 Mon Sep 17 00:00:00 2001 From: Nicholas Piggin Date: Sat, 26 Nov 2022 19:59:22 +1000 Subject: powerpc/qspinlock: paravirt yield to lock owner Waiters spinning on the lock word should yield to the lock owner if the vCPU is preempted. This improves performance when the hypervisor has oversubscribed physical CPUs. Signed-off-by: Nicholas Piggin Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20221126095932.1234527-8-npiggin@gmail.com --- arch/powerpc/lib/qspinlock.c | 99 ++++++++++++++++++++++++++++++++++++++------ 1 file changed, 87 insertions(+), 12 deletions(-) (limited to 'arch/powerpc/lib') diff --git a/arch/powerpc/lib/qspinlock.c b/arch/powerpc/lib/qspinlock.c index 4d74db0e565f..18e21574e6c5 100644 --- a/arch/powerpc/lib/qspinlock.c +++ b/arch/powerpc/lib/qspinlock.c @@ -5,6 +5,7 @@ #include #include #include +#include #define MAX_NODES 4 @@ -24,14 +25,16 @@ static int steal_spins __read_mostly = (1 << 5); static bool maybe_stealers __read_mostly = true; static int head_spins __read_mostly = (1 << 8); +static bool pv_yield_owner __read_mostly = true; + static DEFINE_PER_CPU_ALIGNED(struct qnodes, qnodes); -static __always_inline int get_steal_spins(void) +static __always_inline int get_steal_spins(bool paravirt) { return steal_spins; } -static __always_inline int get_head_spins(void) +static __always_inline int get_head_spins(bool paravirt) { return head_spins; } @@ -46,6 +49,11 @@ static inline int decode_tail_cpu(u32 val) return (val >> _Q_TAIL_CPU_OFFSET) - 1; } +static inline int get_owner_cpu(u32 val) +{ + return (val & _Q_OWNER_CPU_MASK) >> _Q_OWNER_CPU_OFFSET; +} + /* * Try to acquire the lock if it was not already locked. If the tail matches * mytail then clear it, otherwise leave it unchnaged. Return previous value. @@ -150,7 +158,45 @@ static struct qnode *get_tail_qnode(struct qspinlock *lock, u32 val) BUG(); } -static inline bool try_to_steal_lock(struct qspinlock *lock) +static __always_inline void yield_to_locked_owner(struct qspinlock *lock, u32 val, bool paravirt) +{ + int owner; + u32 yield_count; + + BUG_ON(!(val & _Q_LOCKED_VAL)); + + if (!paravirt) + goto relax; + + if (!pv_yield_owner) + goto relax; + + owner = get_owner_cpu(val); + yield_count = yield_count_of(owner); + + if ((yield_count & 1) == 0) + goto relax; /* owner vcpu is running */ + + /* + * Read the lock word after sampling the yield count. On the other side + * there may a wmb because the yield count update is done by the + * hypervisor preemption and the value update by the OS, however this + * ordering might reduce the chance of out of order accesses and + * improve the heuristic. + */ + smp_rmb(); + + if (READ_ONCE(lock->val) == val) { + yield_to_preempted(owner, yield_count); + /* Don't relax if we yielded. Maybe we should? */ + return; + } +relax: + cpu_relax(); +} + + +static __always_inline bool try_to_steal_lock(struct qspinlock *lock, bool paravirt) { int iters = 0; @@ -168,16 +214,16 @@ static inline bool try_to_steal_lock(struct qspinlock *lock) if (__queued_spin_trylock_steal(lock)) return true; } else { - cpu_relax(); + yield_to_locked_owner(lock, val, paravirt); } iters++; - } while (iters < get_steal_spins()); + } while (iters < get_steal_spins(paravirt)); return false; } -static inline void queued_spin_lock_mcs_queue(struct qspinlock *lock) +static __always_inline void queued_spin_lock_mcs_queue(struct qspinlock *lock, bool paravirt) { struct qnodes *qnodesp; struct qnode *next, *node; @@ -235,12 +281,12 @@ again: if (!(val & _Q_LOCKED_VAL)) break; - cpu_relax(); + yield_to_locked_owner(lock, val, paravirt); if (!maybe_stealers) continue; iters++; - if (!mustq && iters >= get_head_spins()) { + if (!mustq && iters >= get_head_spins(paravirt)) { mustq = true; set_mustq(lock); val |= _Q_MUST_Q_VAL; @@ -276,10 +322,20 @@ release: void queued_spin_lock_slowpath(struct qspinlock *lock) { - if (try_to_steal_lock(lock)) - return; - - queued_spin_lock_mcs_queue(lock); + /* + * This looks funny, but it induces the compiler to inline both + * sides of the branch rather than share code as when the condition + * is passed as the paravirt argument to the functions. + */ + if (IS_ENABLED(CONFIG_PARAVIRT_SPINLOCKS) && is_shared_processor()) { + if (try_to_steal_lock(lock, true)) + return; + queued_spin_lock_mcs_queue(lock, true); + } else { + if (try_to_steal_lock(lock, false)) + return; + queued_spin_lock_mcs_queue(lock, false); + } } EXPORT_SYMBOL(queued_spin_lock_slowpath); @@ -345,10 +401,29 @@ static int head_spins_get(void *data, u64 *val) DEFINE_SIMPLE_ATTRIBUTE(fops_head_spins, head_spins_get, head_spins_set, "%llu\n"); +static int pv_yield_owner_set(void *data, u64 val) +{ + pv_yield_owner = !!val; + + return 0; +} + +static int pv_yield_owner_get(void *data, u64 *val) +{ + *val = pv_yield_owner; + + return 0; +} + +DEFINE_SIMPLE_ATTRIBUTE(fops_pv_yield_owner, pv_yield_owner_get, pv_yield_owner_set, "%llu\n"); + static __init int spinlock_debugfs_init(void) { debugfs_create_file("qspl_steal_spins", 0600, arch_debugfs_dir, NULL, &fops_steal_spins); debugfs_create_file("qspl_head_spins", 0600, arch_debugfs_dir, NULL, &fops_head_spins); + if (is_shared_processor()) { + debugfs_create_file("qspl_pv_yield_owner", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_owner); + } return 0; } -- cgit v1.2.3 From bd48287b2cf4cd6e95576db3a94fd2a7cdf9832d Mon Sep 17 00:00:00 2001 From: Nicholas Piggin Date: Sat, 26 Nov 2022 19:59:23 +1000 Subject: powerpc/qspinlock: implement option to yield to previous node Queued waiters which are not at the head of the queue don't spin on the lock word but their qnode lock word, waiting for the previous queued CPU to release them. Add an option which allows these waiters to yield to the previous CPU if its vCPU is preempted. Signed-off-by: Nicholas Piggin Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20221126095932.1234527-9-npiggin@gmail.com --- arch/powerpc/lib/qspinlock.c | 46 +++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 45 insertions(+), 1 deletion(-) (limited to 'arch/powerpc/lib') diff --git a/arch/powerpc/lib/qspinlock.c b/arch/powerpc/lib/qspinlock.c index 18e21574e6c5..41afd8e68918 100644 --- a/arch/powerpc/lib/qspinlock.c +++ b/arch/powerpc/lib/qspinlock.c @@ -26,6 +26,7 @@ static bool maybe_stealers __read_mostly = true; static int head_spins __read_mostly = (1 << 8); static bool pv_yield_owner __read_mostly = true; +static bool pv_yield_prev __read_mostly = true; static DEFINE_PER_CPU_ALIGNED(struct qnodes, qnodes); @@ -195,6 +196,32 @@ relax: cpu_relax(); } +static __always_inline void yield_to_prev(struct qspinlock *lock, struct qnode *node, u32 val, bool paravirt) +{ + int prev_cpu = decode_tail_cpu(val); + u32 yield_count; + + if (!paravirt) + goto relax; + + if (!pv_yield_prev) + goto relax; + + yield_count = yield_count_of(prev_cpu); + if ((yield_count & 1) == 0) + goto relax; /* owner vcpu is running */ + + smp_rmb(); /* See yield_to_locked_owner comment */ + + if (!node->locked) { + yield_to_preempted(prev_cpu, yield_count); + return; + } + +relax: + cpu_relax(); +} + static __always_inline bool try_to_steal_lock(struct qspinlock *lock, bool paravirt) { @@ -269,7 +296,7 @@ static __always_inline void queued_spin_lock_mcs_queue(struct qspinlock *lock, b /* Wait for mcs node lock to be released */ while (!node->locked) - cpu_relax(); + yield_to_prev(lock, node, old, paravirt); smp_rmb(); /* acquire barrier for the mcs lock */ } @@ -417,12 +444,29 @@ static int pv_yield_owner_get(void *data, u64 *val) DEFINE_SIMPLE_ATTRIBUTE(fops_pv_yield_owner, pv_yield_owner_get, pv_yield_owner_set, "%llu\n"); +static int pv_yield_prev_set(void *data, u64 val) +{ + pv_yield_prev = !!val; + + return 0; +} + +static int pv_yield_prev_get(void *data, u64 *val) +{ + *val = pv_yield_prev; + + return 0; +} + +DEFINE_SIMPLE_ATTRIBUTE(fops_pv_yield_prev, pv_yield_prev_get, pv_yield_prev_set, "%llu\n"); + static __init int spinlock_debugfs_init(void) { debugfs_create_file("qspl_steal_spins", 0600, arch_debugfs_dir, NULL, &fops_steal_spins); debugfs_create_file("qspl_head_spins", 0600, arch_debugfs_dir, NULL, &fops_head_spins); if (is_shared_processor()) { debugfs_create_file("qspl_pv_yield_owner", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_owner); + debugfs_create_file("qspl_pv_yield_prev", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_prev); } return 0; -- cgit v1.2.3 From b4c3cdc1a698a2f6168768d0bed4bf062723722e Mon Sep 17 00:00:00 2001 From: Nicholas Piggin Date: Sat, 26 Nov 2022 19:59:24 +1000 Subject: powerpc/qspinlock: allow stealing when head of queue yields If the head of queue is preventing stealing but it finds the owner vCPU is preempted, it will yield its cycles to the owner which could cause it to become preempted. Add an option to re-allow stealers before yielding, and disallow them again after returning from the yield. Disable this option by default for now, i.e., no logical change. Signed-off-by: Nicholas Piggin Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20221126095932.1234527-10-npiggin@gmail.com --- arch/powerpc/lib/qspinlock.c | 59 +++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 56 insertions(+), 3 deletions(-) (limited to 'arch/powerpc/lib') diff --git a/arch/powerpc/lib/qspinlock.c b/arch/powerpc/lib/qspinlock.c index 41afd8e68918..c1f3b699b63f 100644 --- a/arch/powerpc/lib/qspinlock.c +++ b/arch/powerpc/lib/qspinlock.c @@ -26,6 +26,7 @@ static bool maybe_stealers __read_mostly = true; static int head_spins __read_mostly = (1 << 8); static bool pv_yield_owner __read_mostly = true; +static bool pv_yield_allow_steal __read_mostly = false; static bool pv_yield_prev __read_mostly = true; static DEFINE_PER_CPU_ALIGNED(struct qnodes, qnodes); @@ -135,6 +136,22 @@ static __always_inline u32 set_mustq(struct qspinlock *lock) return prev; } +static __always_inline u32 clear_mustq(struct qspinlock *lock) +{ + u32 prev; + + asm volatile( +"1: lwarx %0,0,%1 # clear_mustq \n" +" andc %0,%0,%2 \n" +" stwcx. %0,0,%1 \n" +" bne- 1b \n" + : "=&r" (prev) + : "r" (&lock->val), "r" (_Q_MUST_Q_VAL) + : "cr0", "memory"); + + return prev; +} + static struct qnode *get_tail_qnode(struct qspinlock *lock, u32 val) { int cpu = decode_tail_cpu(val); @@ -159,7 +176,7 @@ static struct qnode *get_tail_qnode(struct qspinlock *lock, u32 val) BUG(); } -static __always_inline void yield_to_locked_owner(struct qspinlock *lock, u32 val, bool paravirt) +static __always_inline void __yield_to_locked_owner(struct qspinlock *lock, u32 val, bool paravirt, bool mustq) { int owner; u32 yield_count; @@ -188,7 +205,11 @@ static __always_inline void yield_to_locked_owner(struct qspinlock *lock, u32 va smp_rmb(); if (READ_ONCE(lock->val) == val) { + if (mustq) + clear_mustq(lock); yield_to_preempted(owner, yield_count); + if (mustq) + set_mustq(lock); /* Don't relax if we yielded. Maybe we should? */ return; } @@ -196,6 +217,21 @@ relax: cpu_relax(); } +static __always_inline void yield_to_locked_owner(struct qspinlock *lock, u32 val, bool paravirt) +{ + __yield_to_locked_owner(lock, val, paravirt, false); +} + +static __always_inline void yield_head_to_locked_owner(struct qspinlock *lock, u32 val, bool paravirt) +{ + bool mustq = false; + + if ((val & _Q_MUST_Q_VAL) && pv_yield_allow_steal) + mustq = true; + + __yield_to_locked_owner(lock, val, paravirt, mustq); +} + static __always_inline void yield_to_prev(struct qspinlock *lock, struct qnode *node, u32 val, bool paravirt) { int prev_cpu = decode_tail_cpu(val); @@ -211,7 +247,7 @@ static __always_inline void yield_to_prev(struct qspinlock *lock, struct qnode * if ((yield_count & 1) == 0) goto relax; /* owner vcpu is running */ - smp_rmb(); /* See yield_to_locked_owner comment */ + smp_rmb(); /* See __yield_to_locked_owner comment */ if (!node->locked) { yield_to_preempted(prev_cpu, yield_count); @@ -308,7 +344,7 @@ again: if (!(val & _Q_LOCKED_VAL)) break; - yield_to_locked_owner(lock, val, paravirt); + yield_head_to_locked_owner(lock, val, paravirt); if (!maybe_stealers) continue; iters++; @@ -444,6 +480,22 @@ static int pv_yield_owner_get(void *data, u64 *val) DEFINE_SIMPLE_ATTRIBUTE(fops_pv_yield_owner, pv_yield_owner_get, pv_yield_owner_set, "%llu\n"); +static int pv_yield_allow_steal_set(void *data, u64 val) +{ + pv_yield_allow_steal = !!val; + + return 0; +} + +static int pv_yield_allow_steal_get(void *data, u64 *val) +{ + *val = pv_yield_allow_steal; + + return 0; +} + +DEFINE_SIMPLE_ATTRIBUTE(fops_pv_yield_allow_steal, pv_yield_allow_steal_get, pv_yield_allow_steal_set, "%llu\n"); + static int pv_yield_prev_set(void *data, u64 val) { pv_yield_prev = !!val; @@ -466,6 +518,7 @@ static __init int spinlock_debugfs_init(void) debugfs_create_file("qspl_head_spins", 0600, arch_debugfs_dir, NULL, &fops_head_spins); if (is_shared_processor()) { debugfs_create_file("qspl_pv_yield_owner", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_owner); + debugfs_create_file("qspl_pv_yield_allow_steal", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_allow_steal); debugfs_create_file("qspl_pv_yield_prev", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_prev); } -- cgit v1.2.3 From 28db61e207ea3890d286cff3141c1ce67346074d Mon Sep 17 00:00:00 2001 From: Nicholas Piggin Date: Sat, 26 Nov 2022 19:59:25 +1000 Subject: powerpc/qspinlock: allow propagation of yield CPU down the queue Having all CPUs poll the lock word for the owner CPU that should be yielded to defeats most of the purpose of using MCS queueing for scalability. Yet it may be desirable for queued waiters to yield to a preempted owner. With this change, queue waiters never sample the owner CPU directly from the lock word. The queue head (which is spinning on the lock) propagates the owner CPU back to the next waiter if it finds the owner has been preempted. That waiter then propagates the owner CPU back to the next waiter, and so on. s390 addresses this problem differenty, by having queued waiters sample the lock word to find the owner at a low frequency. That has the advantage of being simpler, the advantage of propagation is that the lock word never has to be accesed by queued waiters, and the transfer of cache lines to transmit the owner data is only required when lock holder vCPU preemption occurs. Signed-off-by: Nicholas Piggin Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20221126095932.1234527-11-npiggin@gmail.com --- arch/powerpc/lib/qspinlock.c | 79 ++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 79 insertions(+) (limited to 'arch/powerpc/lib') diff --git a/arch/powerpc/lib/qspinlock.c b/arch/powerpc/lib/qspinlock.c index c1f3b699b63f..c45f30c9a19e 100644 --- a/arch/powerpc/lib/qspinlock.c +++ b/arch/powerpc/lib/qspinlock.c @@ -12,6 +12,7 @@ struct qnode { struct qnode *next; struct qspinlock *lock; + int yield_cpu; u8 locked; /* 1 if lock acquired */ }; @@ -28,6 +29,7 @@ static int head_spins __read_mostly = (1 << 8); static bool pv_yield_owner __read_mostly = true; static bool pv_yield_allow_steal __read_mostly = false; static bool pv_yield_prev __read_mostly = true; +static bool pv_yield_propagate_owner __read_mostly = true; static DEFINE_PER_CPU_ALIGNED(struct qnodes, qnodes); @@ -232,14 +234,67 @@ static __always_inline void yield_head_to_locked_owner(struct qspinlock *lock, u __yield_to_locked_owner(lock, val, paravirt, mustq); } +static __always_inline void propagate_yield_cpu(struct qnode *node, u32 val, int *set_yield_cpu, bool paravirt) +{ + struct qnode *next; + int owner; + + if (!paravirt) + return; + if (!pv_yield_propagate_owner) + return; + + owner = get_owner_cpu(val); + if (*set_yield_cpu == owner) + return; + + next = READ_ONCE(node->next); + if (!next) + return; + + if (vcpu_is_preempted(owner)) { + next->yield_cpu = owner; + *set_yield_cpu = owner; + } else if (*set_yield_cpu != -1) { + next->yield_cpu = owner; + *set_yield_cpu = owner; + } +} + static __always_inline void yield_to_prev(struct qspinlock *lock, struct qnode *node, u32 val, bool paravirt) { int prev_cpu = decode_tail_cpu(val); u32 yield_count; + int yield_cpu; if (!paravirt) goto relax; + if (!pv_yield_propagate_owner) + goto yield_prev; + + yield_cpu = READ_ONCE(node->yield_cpu); + if (yield_cpu == -1) { + /* Propagate back the -1 CPU */ + if (node->next && node->next->yield_cpu != -1) + node->next->yield_cpu = yield_cpu; + goto yield_prev; + } + + yield_count = yield_count_of(yield_cpu); + if ((yield_count & 1) == 0) + goto yield_prev; /* owner vcpu is running */ + + smp_rmb(); + + if (yield_cpu == node->yield_cpu) { + if (node->next && node->next->yield_cpu != yield_cpu) + node->next->yield_cpu = yield_cpu; + yield_to_preempted(yield_cpu, yield_count); + return; + } + +yield_prev: if (!pv_yield_prev) goto relax; @@ -293,6 +348,7 @@ static __always_inline void queued_spin_lock_mcs_queue(struct qspinlock *lock, b u32 val, old, tail; bool mustq = false; int idx; + int set_yield_cpu = -1; int iters = 0; BUILD_BUG_ON(CONFIG_NR_CPUS >= (1U << _Q_TAIL_CPU_BITS)); @@ -314,6 +370,7 @@ static __always_inline void queued_spin_lock_mcs_queue(struct qspinlock *lock, b node = &qnodesp->nodes[idx]; node->next = NULL; node->lock = lock; + node->yield_cpu = -1; node->locked = 0; tail = encode_tail_cpu(smp_processor_id()); @@ -334,6 +391,10 @@ static __always_inline void queued_spin_lock_mcs_queue(struct qspinlock *lock, b while (!node->locked) yield_to_prev(lock, node, old, paravirt); + /* Clear out stale propagated yield_cpu */ + if (paravirt && pv_yield_propagate_owner && node->yield_cpu != -1) + node->yield_cpu = -1; + smp_rmb(); /* acquire barrier for the mcs lock */ } @@ -344,6 +405,7 @@ again: if (!(val & _Q_LOCKED_VAL)) break; + propagate_yield_cpu(node, val, &set_yield_cpu, paravirt); yield_head_to_locked_owner(lock, val, paravirt); if (!maybe_stealers) continue; @@ -512,6 +574,22 @@ static int pv_yield_prev_get(void *data, u64 *val) DEFINE_SIMPLE_ATTRIBUTE(fops_pv_yield_prev, pv_yield_prev_get, pv_yield_prev_set, "%llu\n"); +static int pv_yield_propagate_owner_set(void *data, u64 val) +{ + pv_yield_propagate_owner = !!val; + + return 0; +} + +static int pv_yield_propagate_owner_get(void *data, u64 *val) +{ + *val = pv_yield_propagate_owner; + + return 0; +} + +DEFINE_SIMPLE_ATTRIBUTE(fops_pv_yield_propagate_owner, pv_yield_propagate_owner_get, pv_yield_propagate_owner_set, "%llu\n"); + static __init int spinlock_debugfs_init(void) { debugfs_create_file("qspl_steal_spins", 0600, arch_debugfs_dir, NULL, &fops_steal_spins); @@ -520,6 +598,7 @@ static __init int spinlock_debugfs_init(void) debugfs_create_file("qspl_pv_yield_owner", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_owner); debugfs_create_file("qspl_pv_yield_allow_steal", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_allow_steal); debugfs_create_file("qspl_pv_yield_prev", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_prev); + debugfs_create_file("qspl_pv_yield_propagate_owner", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_propagate_owner); } return 0; -- cgit v1.2.3 From be742c573fdafcfa1752642ca1c7aaf08c258128 Mon Sep 17 00:00:00 2001 From: Nicholas Piggin Date: Sat, 26 Nov 2022 19:59:26 +1000 Subject: powerpc/qspinlock: add ability to prod new queue head CPU After the head of the queue acquires the lock, it releases the next waiter in the queue to become the new head. Add an option to prod the new head if its vCPU was preempted. This may only have an effect if queue waiters are yielding. Disable this option by default for now, i.e., no logical change. Signed-off-by: Nicholas Piggin Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20221126095932.1234527-12-npiggin@gmail.com --- arch/powerpc/lib/qspinlock.c | 31 +++++++++++++++++++++++++++++-- 1 file changed, 29 insertions(+), 2 deletions(-) (limited to 'arch/powerpc/lib') diff --git a/arch/powerpc/lib/qspinlock.c b/arch/powerpc/lib/qspinlock.c index c45f30c9a19e..2f6c0bed25ea 100644 --- a/arch/powerpc/lib/qspinlock.c +++ b/arch/powerpc/lib/qspinlock.c @@ -12,6 +12,7 @@ struct qnode { struct qnode *next; struct qspinlock *lock; + int cpu; int yield_cpu; u8 locked; /* 1 if lock acquired */ }; @@ -30,6 +31,7 @@ static bool pv_yield_owner __read_mostly = true; static bool pv_yield_allow_steal __read_mostly = false; static bool pv_yield_prev __read_mostly = true; static bool pv_yield_propagate_owner __read_mostly = true; +static bool pv_prod_head __read_mostly = false; static DEFINE_PER_CPU_ALIGNED(struct qnodes, qnodes); @@ -370,10 +372,11 @@ static __always_inline void queued_spin_lock_mcs_queue(struct qspinlock *lock, b node = &qnodesp->nodes[idx]; node->next = NULL; node->lock = lock; + node->cpu = smp_processor_id(); node->yield_cpu = -1; node->locked = 0; - tail = encode_tail_cpu(smp_processor_id()); + tail = encode_tail_cpu(node->cpu); old = publish_tail_cpu(lock, tail); @@ -439,7 +442,14 @@ again: * this store to locked. The corresponding barrier is the smp_rmb() * acquire barrier for mcs lock, above. */ - WRITE_ONCE(next->locked, 1); + if (paravirt && pv_prod_head) { + int next_cpu = next->cpu; + WRITE_ONCE(next->locked, 1); + if (vcpu_is_preempted(next_cpu)) + prod_cpu(next_cpu); + } else { + WRITE_ONCE(next->locked, 1); + } release: qnodesp->count--; /* release the node */ @@ -590,6 +600,22 @@ static int pv_yield_propagate_owner_get(void *data, u64 *val) DEFINE_SIMPLE_ATTRIBUTE(fops_pv_yield_propagate_owner, pv_yield_propagate_owner_get, pv_yield_propagate_owner_set, "%llu\n"); +static int pv_prod_head_set(void *data, u64 val) +{ + pv_prod_head = !!val; + + return 0; +} + +static int pv_prod_head_get(void *data, u64 *val) +{ + *val = pv_prod_head; + + return 0; +} + +DEFINE_SIMPLE_ATTRIBUTE(fops_pv_prod_head, pv_prod_head_get, pv_prod_head_set, "%llu\n"); + static __init int spinlock_debugfs_init(void) { debugfs_create_file("qspl_steal_spins", 0600, arch_debugfs_dir, NULL, &fops_steal_spins); @@ -599,6 +625,7 @@ static __init int spinlock_debugfs_init(void) debugfs_create_file("qspl_pv_yield_allow_steal", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_allow_steal); debugfs_create_file("qspl_pv_yield_prev", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_prev); debugfs_create_file("qspl_pv_yield_propagate_owner", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_propagate_owner); + debugfs_create_file("qspl_pv_prod_head", 0600, arch_debugfs_dir, NULL, &fops_pv_prod_head); } return 0; -- cgit v1.2.3 From f61ab43cc1a6146d6eef7e0713a452c3677ad13e Mon Sep 17 00:00:00 2001 From: Nicholas Piggin Date: Sat, 26 Nov 2022 19:59:27 +1000 Subject: powerpc/qspinlock: allow lock stealing in trylock and lock fastpath This change allows trylock to steal the lock. It also allows the initial lock attempt to steal the lock rather than bailing out and going to the slow path. This gives trylock more strength: without this a continually-contended lock will never permit a trylock to succeed. With this change, the trylock has a small but non-zero chance. It also gives the lock fastpath most of the benefit of passing the reservation back through to the steal loop in the slow path without the complexity. Signed-off-by: Nicholas Piggin Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20221126095932.1234527-13-npiggin@gmail.com --- arch/powerpc/include/asm/qspinlock.h | 22 ++++++++++++++++++++-- arch/powerpc/lib/qspinlock.c | 9 +++++++++ 2 files changed, 29 insertions(+), 2 deletions(-) (limited to 'arch/powerpc/lib') diff --git a/arch/powerpc/include/asm/qspinlock.h b/arch/powerpc/include/asm/qspinlock.h index 9572a2ef974d..93b1c976db8a 100644 --- a/arch/powerpc/include/asm/qspinlock.h +++ b/arch/powerpc/include/asm/qspinlock.h @@ -6,6 +6,15 @@ #include #include +/* + * The trylock itself may steal. This makes trylocks slightly stronger, and + * might make spin locks slightly more efficient when stealing. + * + * This is compile-time, so if true then there may always be stealers, so the + * nosteal paths become unused. + */ +#define _Q_SPIN_TRY_LOCK_STEAL 1 + static __always_inline int queued_spin_is_locked(struct qspinlock *lock) { return READ_ONCE(lock->val); @@ -27,13 +36,14 @@ static __always_inline u32 queued_spin_encode_locked_val(void) return _Q_LOCKED_VAL | (smp_processor_id() << _Q_OWNER_CPU_OFFSET); } -static __always_inline int queued_spin_trylock(struct qspinlock *lock) +static __always_inline int __queued_spin_trylock_nosteal(struct qspinlock *lock) { u32 new = queued_spin_encode_locked_val(); u32 prev; + /* Trylock succeeds only when unlocked and no queued nodes */ asm volatile( -"1: lwarx %0,0,%1,%3 # queued_spin_trylock \n" +"1: lwarx %0,0,%1,%3 # __queued_spin_trylock_nosteal \n" " cmpwi 0,%0,0 \n" " bne- 2f \n" " stwcx. %2,0,%1 \n" @@ -72,6 +82,14 @@ static __always_inline int __queued_spin_trylock_steal(struct qspinlock *lock) return likely(!(prev & ~_Q_TAIL_CPU_MASK)); } +static __always_inline int queued_spin_trylock(struct qspinlock *lock) +{ + if (!_Q_SPIN_TRY_LOCK_STEAL) + return __queued_spin_trylock_nosteal(lock); + else + return __queued_spin_trylock_steal(lock); +} + void queued_spin_lock_slowpath(struct qspinlock *lock); static __always_inline void queued_spin_lock(struct qspinlock *lock) diff --git a/arch/powerpc/lib/qspinlock.c b/arch/powerpc/lib/qspinlock.c index 2f6c0bed25ea..8e5b8bc3f094 100644 --- a/arch/powerpc/lib/qspinlock.c +++ b/arch/powerpc/lib/qspinlock.c @@ -24,7 +24,11 @@ struct qnodes { /* Tuning parameters */ static int steal_spins __read_mostly = (1 << 5); +#if _Q_SPIN_TRY_LOCK_STEAL == 1 +static const bool maybe_stealers = true; +#else static bool maybe_stealers __read_mostly = true; +#endif static int head_spins __read_mostly = (1 << 8); static bool pv_yield_owner __read_mostly = true; @@ -483,6 +487,10 @@ void pv_spinlocks_init(void) #include static int steal_spins_set(void *data, u64 val) { +#if _Q_SPIN_TRY_LOCK_STEAL == 1 + /* MAYBE_STEAL remains true */ + steal_spins = val; +#else static DEFINE_MUTEX(lock); /* @@ -507,6 +515,7 @@ static int steal_spins_set(void *data, u64 val) steal_spins = val; } mutex_unlock(&lock); +#endif return 0; } -- cgit v1.2.3 From 71c235027ce7940434acd3f553602ad8b5d36469 Mon Sep 17 00:00:00 2001 From: Nicholas Piggin Date: Sat, 26 Nov 2022 19:59:28 +1000 Subject: powerpc/qspinlock: use spin_begin/end API Use the spin_begin/spin_cpu_relax/spin_end APIs in qspinlock, which helps to prevent threads issuing a lot of expensive priority nops which may not have much effect due to immediately executing low then medium priority. Signed-off-by: Nicholas Piggin Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20221126095932.1234527-14-npiggin@gmail.com --- arch/powerpc/lib/qspinlock.c | 39 +++++++++++++++++++++++++++++++++++---- 1 file changed, 35 insertions(+), 4 deletions(-) (limited to 'arch/powerpc/lib') diff --git a/arch/powerpc/lib/qspinlock.c b/arch/powerpc/lib/qspinlock.c index 8e5b8bc3f094..36aff7defda8 100644 --- a/arch/powerpc/lib/qspinlock.c +++ b/arch/powerpc/lib/qspinlock.c @@ -184,6 +184,7 @@ static struct qnode *get_tail_qnode(struct qspinlock *lock, u32 val) BUG(); } +/* Called inside spin_begin() */ static __always_inline void __yield_to_locked_owner(struct qspinlock *lock, u32 val, bool paravirt, bool mustq) { int owner; @@ -203,6 +204,8 @@ static __always_inline void __yield_to_locked_owner(struct qspinlock *lock, u32 if ((yield_count & 1) == 0) goto relax; /* owner vcpu is running */ + spin_end(); + /* * Read the lock word after sampling the yield count. On the other side * there may a wmb because the yield count update is done by the @@ -218,18 +221,22 @@ static __always_inline void __yield_to_locked_owner(struct qspinlock *lock, u32 yield_to_preempted(owner, yield_count); if (mustq) set_mustq(lock); + spin_begin(); /* Don't relax if we yielded. Maybe we should? */ return; } + spin_begin(); relax: - cpu_relax(); + spin_cpu_relax(); } +/* Called inside spin_begin() */ static __always_inline void yield_to_locked_owner(struct qspinlock *lock, u32 val, bool paravirt) { __yield_to_locked_owner(lock, val, paravirt, false); } +/* Called inside spin_begin() */ static __always_inline void yield_head_to_locked_owner(struct qspinlock *lock, u32 val, bool paravirt) { bool mustq = false; @@ -267,6 +274,7 @@ static __always_inline void propagate_yield_cpu(struct qnode *node, u32 val, int } } +/* Called inside spin_begin() */ static __always_inline void yield_to_prev(struct qspinlock *lock, struct qnode *node, u32 val, bool paravirt) { int prev_cpu = decode_tail_cpu(val); @@ -291,14 +299,18 @@ static __always_inline void yield_to_prev(struct qspinlock *lock, struct qnode * if ((yield_count & 1) == 0) goto yield_prev; /* owner vcpu is running */ + spin_end(); + smp_rmb(); if (yield_cpu == node->yield_cpu) { if (node->next && node->next->yield_cpu != yield_cpu) node->next->yield_cpu = yield_cpu; yield_to_preempted(yield_cpu, yield_count); + spin_begin(); return; } + spin_begin(); yield_prev: if (!pv_yield_prev) @@ -308,15 +320,19 @@ yield_prev: if ((yield_count & 1) == 0) goto relax; /* owner vcpu is running */ + spin_end(); + smp_rmb(); /* See __yield_to_locked_owner comment */ if (!node->locked) { yield_to_preempted(prev_cpu, yield_count); + spin_begin(); return; } + spin_begin(); relax: - cpu_relax(); + spin_cpu_relax(); } @@ -328,6 +344,8 @@ static __always_inline bool try_to_steal_lock(struct qspinlock *lock, bool parav return false; /* Attempt to steal the lock */ + spin_begin(); + do { u32 val = READ_ONCE(lock->val); @@ -335,8 +353,10 @@ static __always_inline bool try_to_steal_lock(struct qspinlock *lock, bool parav break; if (unlikely(!(val & _Q_LOCKED_VAL))) { + spin_end(); if (__queued_spin_trylock_steal(lock)) return true; + spin_begin(); } else { yield_to_locked_owner(lock, val, paravirt); } @@ -344,6 +364,8 @@ static __always_inline bool try_to_steal_lock(struct qspinlock *lock, bool parav iters++; } while (iters < get_steal_spins(paravirt)); + spin_end(); + return false; } @@ -395,8 +417,10 @@ static __always_inline void queued_spin_lock_mcs_queue(struct qspinlock *lock, b WRITE_ONCE(prev->next, node); /* Wait for mcs node lock to be released */ + spin_begin(); while (!node->locked) yield_to_prev(lock, node, old, paravirt); + spin_end(); /* Clear out stale propagated yield_cpu */ if (paravirt && pv_yield_propagate_owner && node->yield_cpu != -1) @@ -407,6 +431,7 @@ static __always_inline void queued_spin_lock_mcs_queue(struct qspinlock *lock, b again: /* We're at the head of the waitqueue, wait for the lock. */ + spin_begin(); for (;;) { val = READ_ONCE(lock->val); if (!(val & _Q_LOCKED_VAL)) @@ -424,6 +449,7 @@ again: val |= _Q_MUST_Q_VAL; } } + spin_end(); /* If we're the last queued, must clean up the tail. */ old = trylock_clean_tail(lock, tail); @@ -436,8 +462,13 @@ again: goto release; /* We were the tail, no next. */ /* There is a next, must wait for node->next != NULL (MCS protocol) */ - while (!(next = READ_ONCE(node->next))) - cpu_relax(); + next = READ_ONCE(node->next); + if (!next) { + spin_begin(); + while (!(next = READ_ONCE(node->next))) + cpu_relax(); + spin_end(); + } /* * Unlock the next mcs waiter node. Release barrier is not required -- cgit v1.2.3 From cc79701114154efe79663ba47d9e51aad2ed3c78 Mon Sep 17 00:00:00 2001 From: Nicholas Piggin Date: Sat, 26 Nov 2022 19:59:29 +1000 Subject: powerpc/qspinlock: reduce remote node steal spins Allow for a reduction in the number of times a CPU from a different node than the owner can attempt to steal the lock before queueing. This could bias the transfer behaviour of the lock across the machine and reduce NUMA crossings. Signed-off-by: Nicholas Piggin Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20221126095932.1234527-15-npiggin@gmail.com --- arch/powerpc/lib/qspinlock.c | 43 ++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 40 insertions(+), 3 deletions(-) (limited to 'arch/powerpc/lib') diff --git a/arch/powerpc/lib/qspinlock.c b/arch/powerpc/lib/qspinlock.c index 36aff7defda8..8c6b5ef87118 100644 --- a/arch/powerpc/lib/qspinlock.c +++ b/arch/powerpc/lib/qspinlock.c @@ -4,6 +4,7 @@ #include #include #include +#include #include #include @@ -24,6 +25,7 @@ struct qnodes { /* Tuning parameters */ static int steal_spins __read_mostly = (1 << 5); +static int remote_steal_spins __read_mostly = (1 << 2); #if _Q_SPIN_TRY_LOCK_STEAL == 1 static const bool maybe_stealers = true; #else @@ -44,6 +46,11 @@ static __always_inline int get_steal_spins(bool paravirt) return steal_spins; } +static __always_inline int get_remote_steal_spins(bool paravirt) +{ + return remote_steal_spins; +} + static __always_inline int get_head_spins(bool paravirt) { return head_spins; @@ -335,10 +342,24 @@ relax: spin_cpu_relax(); } +static __always_inline bool steal_break(u32 val, int iters, bool paravirt) +{ + if (iters >= get_steal_spins(paravirt)) + return true; + + if (IS_ENABLED(CONFIG_NUMA) && + (iters >= get_remote_steal_spins(paravirt))) { + int cpu = get_owner_cpu(val); + if (numa_node_id() != cpu_to_node(cpu)) + return true; + } + return false; +} static __always_inline bool try_to_steal_lock(struct qspinlock *lock, bool paravirt) { int iters = 0; + u32 val; if (!steal_spins) return false; @@ -347,8 +368,7 @@ static __always_inline bool try_to_steal_lock(struct qspinlock *lock, bool parav spin_begin(); do { - u32 val = READ_ONCE(lock->val); - + val = READ_ONCE(lock->val); if (val & _Q_MUST_Q_VAL) break; @@ -362,7 +382,7 @@ static __always_inline bool try_to_steal_lock(struct qspinlock *lock, bool parav } iters++; - } while (iters < get_steal_spins(paravirt)); + } while (!steal_break(val, iters, paravirt)); spin_end(); @@ -560,6 +580,22 @@ static int steal_spins_get(void *data, u64 *val) DEFINE_SIMPLE_ATTRIBUTE(fops_steal_spins, steal_spins_get, steal_spins_set, "%llu\n"); +static int remote_steal_spins_set(void *data, u64 val) +{ + remote_steal_spins = val; + + return 0; +} + +static int remote_steal_spins_get(void *data, u64 *val) +{ + *val = remote_steal_spins; + + return 0; +} + +DEFINE_SIMPLE_ATTRIBUTE(fops_remote_steal_spins, remote_steal_spins_get, remote_steal_spins_set, "%llu\n"); + static int head_spins_set(void *data, u64 val) { head_spins = val; @@ -659,6 +695,7 @@ DEFINE_SIMPLE_ATTRIBUTE(fops_pv_prod_head, pv_prod_head_get, pv_prod_head_set, " static __init int spinlock_debugfs_init(void) { debugfs_create_file("qspl_steal_spins", 0600, arch_debugfs_dir, NULL, &fops_steal_spins); + debugfs_create_file("qspl_remote_steal_spins", 0600, arch_debugfs_dir, NULL, &fops_remote_steal_spins); debugfs_create_file("qspl_head_spins", 0600, arch_debugfs_dir, NULL, &fops_head_spins); if (is_shared_processor()) { debugfs_create_file("qspl_pv_yield_owner", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_owner); -- cgit v1.2.3 From 39dfc73596b48bb50cf7e4f3f54e38427dda5b4e Mon Sep 17 00:00:00 2001 From: Nicholas Piggin Date: Sat, 26 Nov 2022 19:59:30 +1000 Subject: powerpc/qspinlock: allow indefinite spinning on a preempted owner Provide an option that holds off queueing indefinitely while the lock owner is preempted. This could reduce queueing latencies for very overcommitted vcpu situations. This is disabled by default. Signed-off-by: Nicholas Piggin Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20221126095932.1234527-16-npiggin@gmail.com --- arch/powerpc/lib/qspinlock.c | 77 +++++++++++++++++++++++++++++++++++--------- 1 file changed, 62 insertions(+), 15 deletions(-) (limited to 'arch/powerpc/lib') diff --git a/arch/powerpc/lib/qspinlock.c b/arch/powerpc/lib/qspinlock.c index 8c6b5ef87118..eeaaecfd5b77 100644 --- a/arch/powerpc/lib/qspinlock.c +++ b/arch/powerpc/lib/qspinlock.c @@ -35,6 +35,7 @@ static int head_spins __read_mostly = (1 << 8); static bool pv_yield_owner __read_mostly = true; static bool pv_yield_allow_steal __read_mostly = false; +static bool pv_spin_on_preempted_owner __read_mostly = false; static bool pv_yield_prev __read_mostly = true; static bool pv_yield_propagate_owner __read_mostly = true; static bool pv_prod_head __read_mostly = false; @@ -191,11 +192,12 @@ static struct qnode *get_tail_qnode(struct qspinlock *lock, u32 val) BUG(); } -/* Called inside spin_begin() */ -static __always_inline void __yield_to_locked_owner(struct qspinlock *lock, u32 val, bool paravirt, bool mustq) +/* Called inside spin_begin(). Returns whether or not the vCPU was preempted. */ +static __always_inline bool __yield_to_locked_owner(struct qspinlock *lock, u32 val, bool paravirt, bool mustq) { int owner; u32 yield_count; + bool preempted = false; BUG_ON(!(val & _Q_LOCKED_VAL)); @@ -213,6 +215,8 @@ static __always_inline void __yield_to_locked_owner(struct qspinlock *lock, u32 spin_end(); + preempted = true; + /* * Read the lock word after sampling the yield count. On the other side * there may a wmb because the yield count update is done by the @@ -229,29 +233,32 @@ static __always_inline void __yield_to_locked_owner(struct qspinlock *lock, u32 if (mustq) set_mustq(lock); spin_begin(); + /* Don't relax if we yielded. Maybe we should? */ - return; + return preempted; } spin_begin(); relax: spin_cpu_relax(); + + return preempted; } -/* Called inside spin_begin() */ -static __always_inline void yield_to_locked_owner(struct qspinlock *lock, u32 val, bool paravirt) +/* Called inside spin_begin(). Returns whether or not the vCPU was preempted. */ +static __always_inline bool yield_to_locked_owner(struct qspinlock *lock, u32 val, bool paravirt) { - __yield_to_locked_owner(lock, val, paravirt, false); + return __yield_to_locked_owner(lock, val, paravirt, false); } -/* Called inside spin_begin() */ -static __always_inline void yield_head_to_locked_owner(struct qspinlock *lock, u32 val, bool paravirt) +/* Called inside spin_begin(). Returns whether or not the vCPU was preempted. */ +static __always_inline bool yield_head_to_locked_owner(struct qspinlock *lock, u32 val, bool paravirt) { bool mustq = false; if ((val & _Q_MUST_Q_VAL) && pv_yield_allow_steal) mustq = true; - __yield_to_locked_owner(lock, val, paravirt, mustq); + return __yield_to_locked_owner(lock, val, paravirt, mustq); } static __always_inline void propagate_yield_cpu(struct qnode *node, u32 val, int *set_yield_cpu, bool paravirt) @@ -361,13 +368,16 @@ static __always_inline bool try_to_steal_lock(struct qspinlock *lock, bool parav int iters = 0; u32 val; - if (!steal_spins) + if (!steal_spins) { + /* XXX: should spin_on_preempted_owner do anything here? */ return false; + } /* Attempt to steal the lock */ spin_begin(); - do { + bool preempted = false; + val = READ_ONCE(lock->val); if (val & _Q_MUST_Q_VAL) break; @@ -378,10 +388,23 @@ static __always_inline bool try_to_steal_lock(struct qspinlock *lock, bool parav return true; spin_begin(); } else { - yield_to_locked_owner(lock, val, paravirt); + preempted = yield_to_locked_owner(lock, val, paravirt); } - iters++; + if (preempted) { + if (!pv_spin_on_preempted_owner) + iters++; + /* + * pv_spin_on_preempted_owner don't increase iters + * while the owner is preempted -- we won't interfere + * with it by definition. This could introduce some + * latency issue if we continually observe preempted + * owners, but hopefully that's a rare corner case of + * a badly oversubscribed system. + */ + } else { + iters++; + } } while (!steal_break(val, iters, paravirt)); spin_end(); @@ -453,15 +476,22 @@ again: /* We're at the head of the waitqueue, wait for the lock. */ spin_begin(); for (;;) { + bool preempted; + val = READ_ONCE(lock->val); if (!(val & _Q_LOCKED_VAL)) break; propagate_yield_cpu(node, val, &set_yield_cpu, paravirt); - yield_head_to_locked_owner(lock, val, paravirt); + preempted = yield_head_to_locked_owner(lock, val, paravirt); if (!maybe_stealers) continue; - iters++; + if (preempted) { + if (!pv_spin_on_preempted_owner) + iters++; + } else { + iters++; + } if (!mustq && iters >= get_head_spins(paravirt)) { mustq = true; @@ -644,6 +674,22 @@ static int pv_yield_allow_steal_get(void *data, u64 *val) DEFINE_SIMPLE_ATTRIBUTE(fops_pv_yield_allow_steal, pv_yield_allow_steal_get, pv_yield_allow_steal_set, "%llu\n"); +static int pv_spin_on_preempted_owner_set(void *data, u64 val) +{ + pv_spin_on_preempted_owner = !!val; + + return 0; +} + +static int pv_spin_on_preempted_owner_get(void *data, u64 *val) +{ + *val = pv_spin_on_preempted_owner; + + return 0; +} + +DEFINE_SIMPLE_ATTRIBUTE(fops_pv_spin_on_preempted_owner, pv_spin_on_preempted_owner_get, pv_spin_on_preempted_owner_set, "%llu\n"); + static int pv_yield_prev_set(void *data, u64 val) { pv_yield_prev = !!val; @@ -700,6 +746,7 @@ static __init int spinlock_debugfs_init(void) if (is_shared_processor()) { debugfs_create_file("qspl_pv_yield_owner", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_owner); debugfs_create_file("qspl_pv_yield_allow_steal", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_allow_steal); + debugfs_create_file("qspl_pv_spin_on_preempted_owner", 0600, arch_debugfs_dir, NULL, &fops_pv_spin_on_preempted_owner); debugfs_create_file("qspl_pv_yield_prev", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_prev); debugfs_create_file("qspl_pv_yield_propagate_owner", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_propagate_owner); debugfs_create_file("qspl_pv_prod_head", 0600, arch_debugfs_dir, NULL, &fops_pv_prod_head); -- cgit v1.2.3 From 12b459a5ebf3308e718bc1dd48acb7c4cf7f1a75 Mon Sep 17 00:00:00 2001 From: Nicholas Piggin Date: Sat, 26 Nov 2022 19:59:31 +1000 Subject: powerpc/qspinlock: provide accounting and options for sleepy locks Finding the owner or a queued waiter on a lock with a preempted vcpu is indicative of an oversubscribed guest causing the lock to get into trouble. Provide some options to detect this situation and have new CPUs avoid queueing for a longer time (more steal iterations) to minimise the problems caused by vcpu preemption on the queue. Signed-off-by: Nicholas Piggin Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20221126095932.1234527-17-npiggin@gmail.com --- arch/powerpc/include/asm/qspinlock_types.h | 7 +- arch/powerpc/lib/qspinlock.c | 242 ++++++++++++++++++++++++++--- 2 files changed, 230 insertions(+), 19 deletions(-) (limited to 'arch/powerpc/lib') diff --git a/arch/powerpc/include/asm/qspinlock_types.h b/arch/powerpc/include/asm/qspinlock_types.h index adfeed4aa495..4766a7aa03cb 100644 --- a/arch/powerpc/include/asm/qspinlock_types.h +++ b/arch/powerpc/include/asm/qspinlock_types.h @@ -30,7 +30,7 @@ typedef struct qspinlock { * * 0: locked bit * 1-14: lock holder cpu - * 15: unused bit + * 15: lock owner or queuer vcpus observed to be preempted bit * 16: must queue bit * 17-31: tail cpu (+1) */ @@ -50,6 +50,11 @@ typedef struct qspinlock { #error "qspinlock does not support such large CONFIG_NR_CPUS" #endif +/* 0x00008000 */ +#define _Q_SLEEPY_OFFSET 15 +#define _Q_SLEEPY_BITS 1 +#define _Q_SLEEPY_VAL (1U << _Q_SLEEPY_OFFSET) + /* 0x00010000 */ #define _Q_MUST_Q_OFFSET 16 #define _Q_MUST_Q_BITS 1 diff --git a/arch/powerpc/lib/qspinlock.c b/arch/powerpc/lib/qspinlock.c index eeaaecfd5b77..0f33a07c1d19 100644 --- a/arch/powerpc/lib/qspinlock.c +++ b/arch/powerpc/lib/qspinlock.c @@ -5,6 +5,7 @@ #include #include #include +#include #include #include @@ -36,25 +37,56 @@ static int head_spins __read_mostly = (1 << 8); static bool pv_yield_owner __read_mostly = true; static bool pv_yield_allow_steal __read_mostly = false; static bool pv_spin_on_preempted_owner __read_mostly = false; +static bool pv_sleepy_lock __read_mostly = true; +static bool pv_sleepy_lock_sticky __read_mostly = false; +static u64 pv_sleepy_lock_interval_ns __read_mostly = 0; +static int pv_sleepy_lock_factor __read_mostly = 256; static bool pv_yield_prev __read_mostly = true; static bool pv_yield_propagate_owner __read_mostly = true; static bool pv_prod_head __read_mostly = false; static DEFINE_PER_CPU_ALIGNED(struct qnodes, qnodes); +static DEFINE_PER_CPU_ALIGNED(u64, sleepy_lock_seen_clock); -static __always_inline int get_steal_spins(bool paravirt) +static __always_inline bool recently_sleepy(void) { - return steal_spins; + /* pv_sleepy_lock is true when this is called */ + if (pv_sleepy_lock_interval_ns) { + u64 seen = this_cpu_read(sleepy_lock_seen_clock); + + if (seen) { + u64 delta = sched_clock() - seen; + if (delta < pv_sleepy_lock_interval_ns) + return true; + this_cpu_write(sleepy_lock_seen_clock, 0); + } + } + + return false; } -static __always_inline int get_remote_steal_spins(bool paravirt) +static __always_inline int get_steal_spins(bool paravirt, bool sleepy) { - return remote_steal_spins; + if (paravirt && sleepy) + return steal_spins * pv_sleepy_lock_factor; + else + return steal_spins; } -static __always_inline int get_head_spins(bool paravirt) +static __always_inline int get_remote_steal_spins(bool paravirt, bool sleepy) { - return head_spins; + if (paravirt && sleepy) + return remote_steal_spins * pv_sleepy_lock_factor; + else + return remote_steal_spins; +} + +static __always_inline int get_head_spins(bool paravirt, bool sleepy) +{ + if (paravirt && sleepy) + return head_spins * pv_sleepy_lock_factor; + else + return head_spins; } static inline u32 encode_tail_cpu(int cpu) @@ -168,6 +200,56 @@ static __always_inline u32 clear_mustq(struct qspinlock *lock) return prev; } +static __always_inline bool try_set_sleepy(struct qspinlock *lock, u32 old) +{ + u32 prev; + u32 new = old | _Q_SLEEPY_VAL; + + BUG_ON(!(old & _Q_LOCKED_VAL)); + BUG_ON(old & _Q_SLEEPY_VAL); + + asm volatile( +"1: lwarx %0,0,%1 # try_set_sleepy \n" +" cmpw 0,%0,%2 \n" +" bne- 2f \n" +" stwcx. %3,0,%1 \n" +" bne- 1b \n" +"2: \n" + : "=&r" (prev) + : "r" (&lock->val), "r"(old), "r" (new) + : "cr0", "memory"); + + return likely(prev == old); +} + +static __always_inline void seen_sleepy_owner(struct qspinlock *lock, u32 val) +{ + if (pv_sleepy_lock) { + if (pv_sleepy_lock_interval_ns) + this_cpu_write(sleepy_lock_seen_clock, sched_clock()); + if (!(val & _Q_SLEEPY_VAL)) + try_set_sleepy(lock, val); + } +} + +static __always_inline void seen_sleepy_lock(void) +{ + if (pv_sleepy_lock && pv_sleepy_lock_interval_ns) + this_cpu_write(sleepy_lock_seen_clock, sched_clock()); +} + +static __always_inline void seen_sleepy_node(struct qspinlock *lock, u32 val) +{ + if (pv_sleepy_lock) { + if (pv_sleepy_lock_interval_ns) + this_cpu_write(sleepy_lock_seen_clock, sched_clock()); + if (val & _Q_LOCKED_VAL) { + if (!(val & _Q_SLEEPY_VAL)) + try_set_sleepy(lock, val); + } + } +} + static struct qnode *get_tail_qnode(struct qspinlock *lock, u32 val) { int cpu = decode_tail_cpu(val); @@ -215,6 +297,7 @@ static __always_inline bool __yield_to_locked_owner(struct qspinlock *lock, u32 spin_end(); + seen_sleepy_owner(lock, val); preempted = true; /* @@ -289,11 +372,12 @@ static __always_inline void propagate_yield_cpu(struct qnode *node, u32 val, int } /* Called inside spin_begin() */ -static __always_inline void yield_to_prev(struct qspinlock *lock, struct qnode *node, u32 val, bool paravirt) +static __always_inline bool yield_to_prev(struct qspinlock *lock, struct qnode *node, u32 val, bool paravirt) { int prev_cpu = decode_tail_cpu(val); u32 yield_count; int yield_cpu; + bool preempted = false; if (!paravirt) goto relax; @@ -315,6 +399,9 @@ static __always_inline void yield_to_prev(struct qspinlock *lock, struct qnode * spin_end(); + preempted = true; + seen_sleepy_node(lock, val); + smp_rmb(); if (yield_cpu == node->yield_cpu) { @@ -322,7 +409,7 @@ static __always_inline void yield_to_prev(struct qspinlock *lock, struct qnode * node->next->yield_cpu = yield_cpu; yield_to_preempted(yield_cpu, yield_count); spin_begin(); - return; + return preempted; } spin_begin(); @@ -336,26 +423,31 @@ yield_prev: spin_end(); + preempted = true; + seen_sleepy_node(lock, val); + smp_rmb(); /* See __yield_to_locked_owner comment */ if (!node->locked) { yield_to_preempted(prev_cpu, yield_count); spin_begin(); - return; + return preempted; } spin_begin(); relax: spin_cpu_relax(); + + return preempted; } -static __always_inline bool steal_break(u32 val, int iters, bool paravirt) +static __always_inline bool steal_break(u32 val, int iters, bool paravirt, bool sleepy) { - if (iters >= get_steal_spins(paravirt)) + if (iters >= get_steal_spins(paravirt, sleepy)) return true; if (IS_ENABLED(CONFIG_NUMA) && - (iters >= get_remote_steal_spins(paravirt))) { + (iters >= get_remote_steal_spins(paravirt, sleepy))) { int cpu = get_owner_cpu(val); if (numa_node_id() != cpu_to_node(cpu)) return true; @@ -365,6 +457,8 @@ static __always_inline bool steal_break(u32 val, int iters, bool paravirt) static __always_inline bool try_to_steal_lock(struct qspinlock *lock, bool paravirt) { + bool seen_preempted = false; + bool sleepy = false; int iters = 0; u32 val; @@ -391,7 +485,25 @@ static __always_inline bool try_to_steal_lock(struct qspinlock *lock, bool parav preempted = yield_to_locked_owner(lock, val, paravirt); } + if (paravirt && pv_sleepy_lock) { + if (!sleepy) { + if (val & _Q_SLEEPY_VAL) { + seen_sleepy_lock(); + sleepy = true; + } else if (recently_sleepy()) { + sleepy = true; + } + } + if (pv_sleepy_lock_sticky && seen_preempted && + !(val & _Q_SLEEPY_VAL)) { + if (try_set_sleepy(lock, val)) + val |= _Q_SLEEPY_VAL; + } + } + if (preempted) { + seen_preempted = true; + sleepy = true; if (!pv_spin_on_preempted_owner) iters++; /* @@ -405,7 +517,7 @@ static __always_inline bool try_to_steal_lock(struct qspinlock *lock, bool parav } else { iters++; } - } while (!steal_break(val, iters, paravirt)); + } while (!steal_break(val, iters, paravirt, sleepy)); spin_end(); @@ -417,6 +529,8 @@ static __always_inline void queued_spin_lock_mcs_queue(struct qspinlock *lock, b struct qnodes *qnodesp; struct qnode *next, *node; u32 val, old, tail; + bool seen_preempted = false; + bool sleepy = false; bool mustq = false; int idx; int set_yield_cpu = -1; @@ -461,8 +575,10 @@ static __always_inline void queued_spin_lock_mcs_queue(struct qspinlock *lock, b /* Wait for mcs node lock to be released */ spin_begin(); - while (!node->locked) - yield_to_prev(lock, node, old, paravirt); + while (!node->locked) { + if (yield_to_prev(lock, node, old, paravirt)) + seen_preempted = true; + } spin_end(); /* Clear out stale propagated yield_cpu */ @@ -472,8 +588,8 @@ static __always_inline void queued_spin_lock_mcs_queue(struct qspinlock *lock, b smp_rmb(); /* acquire barrier for the mcs lock */ } -again: /* We're at the head of the waitqueue, wait for the lock. */ +again: spin_begin(); for (;;) { bool preempted; @@ -482,18 +598,40 @@ again: if (!(val & _Q_LOCKED_VAL)) break; + if (paravirt && pv_sleepy_lock && maybe_stealers) { + if (!sleepy) { + if (val & _Q_SLEEPY_VAL) { + seen_sleepy_lock(); + sleepy = true; + } else if (recently_sleepy()) { + sleepy = true; + } + } + if (pv_sleepy_lock_sticky && seen_preempted && + !(val & _Q_SLEEPY_VAL)) { + if (try_set_sleepy(lock, val)) + val |= _Q_SLEEPY_VAL; + } + } + propagate_yield_cpu(node, val, &set_yield_cpu, paravirt); preempted = yield_head_to_locked_owner(lock, val, paravirt); if (!maybe_stealers) continue; - if (preempted) { + + if (preempted) + seen_preempted = true; + + if (paravirt && preempted) { + sleepy = true; + if (!pv_spin_on_preempted_owner) iters++; } else { iters++; } - if (!mustq && iters >= get_head_spins(paravirt)) { + if (!mustq && iters >= get_head_spins(paravirt, sleepy)) { mustq = true; set_mustq(lock); val |= _Q_MUST_Q_VAL; @@ -690,6 +828,70 @@ static int pv_spin_on_preempted_owner_get(void *data, u64 *val) DEFINE_SIMPLE_ATTRIBUTE(fops_pv_spin_on_preempted_owner, pv_spin_on_preempted_owner_get, pv_spin_on_preempted_owner_set, "%llu\n"); +static int pv_sleepy_lock_set(void *data, u64 val) +{ + pv_sleepy_lock = !!val; + + return 0; +} + +static int pv_sleepy_lock_get(void *data, u64 *val) +{ + *val = pv_sleepy_lock; + + return 0; +} + +DEFINE_SIMPLE_ATTRIBUTE(fops_pv_sleepy_lock, pv_sleepy_lock_get, pv_sleepy_lock_set, "%llu\n"); + +static int pv_sleepy_lock_sticky_set(void *data, u64 val) +{ + pv_sleepy_lock_sticky = !!val; + + return 0; +} + +static int pv_sleepy_lock_sticky_get(void *data, u64 *val) +{ + *val = pv_sleepy_lock_sticky; + + return 0; +} + +DEFINE_SIMPLE_ATTRIBUTE(fops_pv_sleepy_lock_sticky, pv_sleepy_lock_sticky_get, pv_sleepy_lock_sticky_set, "%llu\n"); + +static int pv_sleepy_lock_interval_ns_set(void *data, u64 val) +{ + pv_sleepy_lock_interval_ns = val; + + return 0; +} + +static int pv_sleepy_lock_interval_ns_get(void *data, u64 *val) +{ + *val = pv_sleepy_lock_interval_ns; + + return 0; +} + +DEFINE_SIMPLE_ATTRIBUTE(fops_pv_sleepy_lock_interval_ns, pv_sleepy_lock_interval_ns_get, pv_sleepy_lock_interval_ns_set, "%llu\n"); + +static int pv_sleepy_lock_factor_set(void *data, u64 val) +{ + pv_sleepy_lock_factor = val; + + return 0; +} + +static int pv_sleepy_lock_factor_get(void *data, u64 *val) +{ + *val = pv_sleepy_lock_factor; + + return 0; +} + +DEFINE_SIMPLE_ATTRIBUTE(fops_pv_sleepy_lock_factor, pv_sleepy_lock_factor_get, pv_sleepy_lock_factor_set, "%llu\n"); + static int pv_yield_prev_set(void *data, u64 val) { pv_yield_prev = !!val; @@ -747,6 +949,10 @@ static __init int spinlock_debugfs_init(void) debugfs_create_file("qspl_pv_yield_owner", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_owner); debugfs_create_file("qspl_pv_yield_allow_steal", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_allow_steal); debugfs_create_file("qspl_pv_spin_on_preempted_owner", 0600, arch_debugfs_dir, NULL, &fops_pv_spin_on_preempted_owner); + debugfs_create_file("qspl_pv_sleepy_lock", 0600, arch_debugfs_dir, NULL, &fops_pv_sleepy_lock); + debugfs_create_file("qspl_pv_sleepy_lock_sticky", 0600, arch_debugfs_dir, NULL, &fops_pv_sleepy_lock_sticky); + debugfs_create_file("qspl_pv_sleepy_lock_interval_ns", 0600, arch_debugfs_dir, NULL, &fops_pv_sleepy_lock_interval_ns); + debugfs_create_file("qspl_pv_sleepy_lock_factor", 0600, arch_debugfs_dir, NULL, &fops_pv_sleepy_lock_factor); debugfs_create_file("qspl_pv_yield_prev", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_prev); debugfs_create_file("qspl_pv_yield_propagate_owner", 0600, arch_debugfs_dir, NULL, &fops_pv_yield_propagate_owner); debugfs_create_file("qspl_pv_prod_head", 0600, arch_debugfs_dir, NULL, &fops_pv_prod_head); -- cgit v1.2.3 From 0b2199841a7952d01a717b465df028b40b2cf3e9 Mon Sep 17 00:00:00 2001 From: Nicholas Piggin Date: Sat, 26 Nov 2022 19:59:32 +1000 Subject: powerpc/qspinlock: add compile-time tuning adjustments This adds compile-time options that allow the EH lock hint bit to be enabled or disabled, and adds some new options that may or may not help matters. To help with experimentation and tuning. Signed-off-by: Nicholas Piggin Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20221126095932.1234527-18-npiggin@gmail.com --- arch/powerpc/include/asm/qspinlock.h | 61 ++++++++++++++++++++++++++++++++++-- arch/powerpc/lib/qspinlock.c | 39 +++++++++++++++++++++-- 2 files changed, 94 insertions(+), 6 deletions(-) (limited to 'arch/powerpc/lib') diff --git a/arch/powerpc/include/asm/qspinlock.h b/arch/powerpc/include/asm/qspinlock.h index 93b1c976db8a..28a53fb69b38 100644 --- a/arch/powerpc/include/asm/qspinlock.h +++ b/arch/powerpc/include/asm/qspinlock.h @@ -6,15 +6,68 @@ #include #include +#ifdef CONFIG_PPC64 +/* + * Use the EH=1 hint for accesses that result in the lock being acquired. + * The hardware is supposed to optimise this pattern by holding the lock + * cacheline longer, and releasing when a store to the same memory (the + * unlock) is performed. + */ +#define _Q_SPIN_EH_HINT 1 +#else +#define _Q_SPIN_EH_HINT 0 +#endif + /* * The trylock itself may steal. This makes trylocks slightly stronger, and - * might make spin locks slightly more efficient when stealing. + * makes locks slightly more efficient when stealing. * * This is compile-time, so if true then there may always be stealers, so the * nosteal paths become unused. */ #define _Q_SPIN_TRY_LOCK_STEAL 1 +/* + * Put a speculation barrier after testing the lock/node and finding it + * busy. Try to prevent pointless speculation in slow paths. + * + * Slows down the lockstorm microbenchmark with no stealing, where locking + * is purely FIFO through the queue. May have more benefit in real workload + * where speculating into the wrong place could have a greater cost. + */ +#define _Q_SPIN_SPEC_BARRIER 0 + +#ifdef CONFIG_PPC64 +/* + * Execute a miso instruction after passing the MCS lock ownership to the + * queue head. Miso is intended to make stores visible to other CPUs sooner. + * + * This seems to make the lockstorm microbenchmark nospin test go slightly + * faster on POWER10, but disable for now. + */ +#define _Q_SPIN_MISO 0 +#else +#define _Q_SPIN_MISO 0 +#endif + +#ifdef CONFIG_PPC64 +/* + * This executes miso after an unlock of the lock word, having ownership + * pass to the next CPU sooner. This will slow the uncontended path to some + * degree. Not evidence it helps yet. + */ +#define _Q_SPIN_MISO_UNLOCK 0 +#else +#define _Q_SPIN_MISO_UNLOCK 0 +#endif + +/* + * Seems to slow down lockstorm microbenchmark, suspect queue node just + * has to become shared again right afterwards when its waiter spins on + * the lock field. + */ +#define _Q_SPIN_PREFETCH_NEXT 0 + static __always_inline int queued_spin_is_locked(struct qspinlock *lock) { return READ_ONCE(lock->val); @@ -52,7 +105,7 @@ static __always_inline int __queued_spin_trylock_nosteal(struct qspinlock *lock) "2: \n" : "=&r" (prev) : "r" (&lock->val), "r" (new), - "i" (IS_ENABLED(CONFIG_PPC64)) + "i" (_Q_SPIN_EH_HINT) : "cr0", "memory"); return likely(prev == 0); @@ -76,7 +129,7 @@ static __always_inline int __queued_spin_trylock_steal(struct qspinlock *lock) "2: \n" : "=&r" (prev), "=&r" (tmp) : "r" (&lock->val), "r" (new), "r" (_Q_TAIL_CPU_MASK), - "i" (IS_ENABLED(CONFIG_PPC64)) + "i" (_Q_SPIN_EH_HINT) : "cr0", "memory"); return likely(!(prev & ~_Q_TAIL_CPU_MASK)); @@ -101,6 +154,8 @@ static __always_inline void queued_spin_lock(struct qspinlock *lock) static inline void queued_spin_unlock(struct qspinlock *lock) { smp_store_release(&lock->locked, 0); + if (_Q_SPIN_MISO_UNLOCK) + asm volatile("miso" ::: "memory"); } #define arch_spin_is_locked(l) queued_spin_is_locked(l) diff --git a/arch/powerpc/lib/qspinlock.c b/arch/powerpc/lib/qspinlock.c index 0f33a07c1d19..1cf5d3e75250 100644 --- a/arch/powerpc/lib/qspinlock.c +++ b/arch/powerpc/lib/qspinlock.c @@ -48,6 +48,12 @@ static bool pv_prod_head __read_mostly = false; static DEFINE_PER_CPU_ALIGNED(struct qnodes, qnodes); static DEFINE_PER_CPU_ALIGNED(u64, sleepy_lock_seen_clock); +#if _Q_SPIN_SPEC_BARRIER == 1 +#define spec_barrier() do { asm volatile("ori 31,31,0" ::: "memory"); } while (0) +#else +#define spec_barrier() do { } while (0) +#endif + static __always_inline bool recently_sleepy(void) { /* pv_sleepy_lock is true when this is called */ @@ -137,7 +143,7 @@ static __always_inline u32 trylock_clean_tail(struct qspinlock *lock, u32 tail) : "r" (&lock->val), "r"(tail), "r" (newval), "i" (_Q_LOCKED_VAL), "r" (_Q_TAIL_CPU_MASK), - "i" (IS_ENABLED(CONFIG_PPC64)) + "i" (_Q_SPIN_EH_HINT) : "cr0", "memory"); return prev; @@ -475,6 +481,7 @@ static __always_inline bool try_to_steal_lock(struct qspinlock *lock, bool parav val = READ_ONCE(lock->val); if (val & _Q_MUST_Q_VAL) break; + spec_barrier(); if (unlikely(!(val & _Q_LOCKED_VAL))) { spin_end(); @@ -540,6 +547,7 @@ static __always_inline void queued_spin_lock_mcs_queue(struct qspinlock *lock, b qnodesp = this_cpu_ptr(&qnodes); if (unlikely(qnodesp->count >= MAX_NODES)) { + spec_barrier(); while (!queued_spin_trylock(lock)) cpu_relax(); return; @@ -576,9 +584,12 @@ static __always_inline void queued_spin_lock_mcs_queue(struct qspinlock *lock, b /* Wait for mcs node lock to be released */ spin_begin(); while (!node->locked) { + spec_barrier(); + if (yield_to_prev(lock, node, old, paravirt)) seen_preempted = true; } + spec_barrier(); spin_end(); /* Clear out stale propagated yield_cpu */ @@ -586,6 +597,17 @@ static __always_inline void queued_spin_lock_mcs_queue(struct qspinlock *lock, b node->yield_cpu = -1; smp_rmb(); /* acquire barrier for the mcs lock */ + + /* + * Generic qspinlocks have this prefetch here, but it seems + * like it could cause additional line transitions because + * the waiter will keep loading from it. + */ + if (_Q_SPIN_PREFETCH_NEXT) { + next = READ_ONCE(node->next); + if (next) + prefetchw(next); + } } /* We're at the head of the waitqueue, wait for the lock. */ @@ -597,6 +619,7 @@ again: val = READ_ONCE(lock->val); if (!(val & _Q_LOCKED_VAL)) break; + spec_barrier(); if (paravirt && pv_sleepy_lock && maybe_stealers) { if (!sleepy) { @@ -637,6 +660,7 @@ again: val |= _Q_MUST_Q_VAL; } } + spec_barrier(); spin_end(); /* If we're the last queued, must clean up the tail. */ @@ -657,6 +681,7 @@ again: cpu_relax(); spin_end(); } + spec_barrier(); /* * Unlock the next mcs waiter node. Release barrier is not required @@ -668,10 +693,14 @@ again: if (paravirt && pv_prod_head) { int next_cpu = next->cpu; WRITE_ONCE(next->locked, 1); + if (_Q_SPIN_MISO) + asm volatile("miso" ::: "memory"); if (vcpu_is_preempted(next_cpu)) prod_cpu(next_cpu); } else { WRITE_ONCE(next->locked, 1); + if (_Q_SPIN_MISO) + asm volatile("miso" ::: "memory"); } release: @@ -686,12 +715,16 @@ void queued_spin_lock_slowpath(struct qspinlock *lock) * is passed as the paravirt argument to the functions. */ if (IS_ENABLED(CONFIG_PARAVIRT_SPINLOCKS) && is_shared_processor()) { - if (try_to_steal_lock(lock, true)) + if (try_to_steal_lock(lock, true)) { + spec_barrier(); return; + } queued_spin_lock_mcs_queue(lock, true); } else { - if (try_to_steal_lock(lock, false)) + if (try_to_steal_lock(lock, false)) { + spec_barrier(); return; + } queued_spin_lock_mcs_queue(lock, false); } } -- cgit v1.2.3