^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1) /* SPDX-License-Identifier: GPL-2.0 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2) #ifndef _GEN_PV_LOCK_SLOWPATH
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3) #error "do not include this file"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6) #include <linux/hash.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7) #include <linux/memblock.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8) #include <linux/debug_locks.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11) * Implement paravirt qspinlocks; the general idea is to halt the vcpus instead
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 12) * of spinning them.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 13) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 14) * This relies on the architecture to provide two paravirt hypercalls:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 15) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 16) * pv_wait(u8 *ptr, u8 val) -- suspends the vcpu if *ptr == val
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 17) * pv_kick(cpu) -- wakes a suspended vcpu
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 18) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 19) * Using these we implement __pv_queued_spin_lock_slowpath() and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 20) * __pv_queued_spin_unlock() to replace native_queued_spin_lock_slowpath() and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 21) * native_queued_spin_unlock().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 22) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 23)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 24) #define _Q_SLOW_VAL (3U << _Q_LOCKED_OFFSET)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 25)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 26) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 27) * Queue Node Adaptive Spinning
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 28) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 29) * A queue node vCPU will stop spinning if the vCPU in the previous node is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 30) * not running. The one lock stealing attempt allowed at slowpath entry
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 31) * mitigates the slight slowdown for non-overcommitted guest with this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 32) * aggressive wait-early mechanism.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 33) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 34) * The status of the previous node will be checked at fixed interval
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 35) * controlled by PV_PREV_CHECK_MASK. This is to ensure that we won't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 36) * pound on the cacheline of the previous node too heavily.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 37) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 38) #define PV_PREV_CHECK_MASK 0xff
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 39)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 40) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 41) * Queue node uses: vcpu_running & vcpu_halted.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 42) * Queue head uses: vcpu_running & vcpu_hashed.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 43) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 44) enum vcpu_state {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 45) vcpu_running = 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 46) vcpu_halted, /* Used only in pv_wait_node */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 47) vcpu_hashed, /* = pv_hash'ed + vcpu_halted */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 48) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 49)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 50) struct pv_node {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 51) struct mcs_spinlock mcs;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 52) int cpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 53) u8 state;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 54) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 55)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 56) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 57) * Hybrid PV queued/unfair lock
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 58) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 59) * By replacing the regular queued_spin_trylock() with the function below,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 60) * it will be called once when a lock waiter enter the PV slowpath before
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 61) * being queued.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 62) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 63) * The pending bit is set by the queue head vCPU of the MCS wait queue in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 64) * pv_wait_head_or_lock() to signal that it is ready to spin on the lock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 65) * When that bit becomes visible to the incoming waiters, no lock stealing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 66) * is allowed. The function will return immediately to make the waiters
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 67) * enter the MCS wait queue. So lock starvation shouldn't happen as long
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 68) * as the queued mode vCPUs are actively running to set the pending bit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 69) * and hence disabling lock stealing.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 70) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 71) * When the pending bit isn't set, the lock waiters will stay in the unfair
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 72) * mode spinning on the lock unless the MCS wait queue is empty. In this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 73) * case, the lock waiters will enter the queued mode slowpath trying to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 74) * become the queue head and set the pending bit.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 75) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 76) * This hybrid PV queued/unfair lock combines the best attributes of a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 77) * queued lock (no lock starvation) and an unfair lock (good performance
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 78) * on not heavily contended locks).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 79) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 80) #define queued_spin_trylock(l) pv_hybrid_queued_unfair_trylock(l)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 81) static inline bool pv_hybrid_queued_unfair_trylock(struct qspinlock *lock)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 82) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 83) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 84) * Stay in unfair lock mode as long as queued mode waiters are
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 85) * present in the MCS wait queue but the pending bit isn't set.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 86) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 87) for (;;) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 88) int val = atomic_read(&lock->val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 89)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 90) if (!(val & _Q_LOCKED_PENDING_MASK) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 91) (cmpxchg_acquire(&lock->locked, 0, _Q_LOCKED_VAL) == 0)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 92) lockevent_inc(pv_lock_stealing);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 93) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 94) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 95) if (!(val & _Q_TAIL_MASK) || (val & _Q_PENDING_MASK))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 96) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 97)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 98) cpu_relax();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 99) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 100)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 101) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 102) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 103)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 104) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 105) * The pending bit is used by the queue head vCPU to indicate that it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 106) * is actively spinning on the lock and no lock stealing is allowed.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 107) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 108) #if _Q_PENDING_BITS == 8
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 109) static __always_inline void set_pending(struct qspinlock *lock)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 110) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 111) WRITE_ONCE(lock->pending, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 112) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 113)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 114) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 115) * The pending bit check in pv_queued_spin_steal_lock() isn't a memory
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 116) * barrier. Therefore, an atomic cmpxchg_acquire() is used to acquire the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 117) * lock just to be sure that it will get it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 118) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 119) static __always_inline int trylock_clear_pending(struct qspinlock *lock)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 120) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 121) return !READ_ONCE(lock->locked) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 122) (cmpxchg_acquire(&lock->locked_pending, _Q_PENDING_VAL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 123) _Q_LOCKED_VAL) == _Q_PENDING_VAL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 124) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 125) #else /* _Q_PENDING_BITS == 8 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 126) static __always_inline void set_pending(struct qspinlock *lock)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 127) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 128) atomic_or(_Q_PENDING_VAL, &lock->val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 129) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 130)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 131) static __always_inline int trylock_clear_pending(struct qspinlock *lock)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 132) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 133) int val = atomic_read(&lock->val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 134)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 135) for (;;) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 136) int old, new;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 137)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 138) if (val & _Q_LOCKED_MASK)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 139) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 140)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 141) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 142) * Try to clear pending bit & set locked bit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 143) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 144) old = val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 145) new = (val & ~_Q_PENDING_MASK) | _Q_LOCKED_VAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 146) val = atomic_cmpxchg_acquire(&lock->val, old, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 147)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 148) if (val == old)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 149) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 150) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 151) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 152) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 153) #endif /* _Q_PENDING_BITS == 8 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 154)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 155) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 156) * Lock and MCS node addresses hash table for fast lookup
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 157) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 158) * Hashing is done on a per-cacheline basis to minimize the need to access
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 159) * more than one cacheline.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 160) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 161) * Dynamically allocate a hash table big enough to hold at least 4X the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 162) * number of possible cpus in the system. Allocation is done on page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 163) * granularity. So the minimum number of hash buckets should be at least
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 164) * 256 (64-bit) or 512 (32-bit) to fully utilize a 4k page.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 165) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 166) * Since we should not be holding locks from NMI context (very rare indeed) the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 167) * max load factor is 0.75, which is around the point where open addressing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 168) * breaks down.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 169) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 170) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 171) struct pv_hash_entry {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 172) struct qspinlock *lock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 173) struct pv_node *node;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 174) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 175)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 176) #define PV_HE_PER_LINE (SMP_CACHE_BYTES / sizeof(struct pv_hash_entry))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 177) #define PV_HE_MIN (PAGE_SIZE / sizeof(struct pv_hash_entry))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 178)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 179) static struct pv_hash_entry *pv_lock_hash;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 180) static unsigned int pv_lock_hash_bits __read_mostly;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 181)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 182) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 183) * Allocate memory for the PV qspinlock hash buckets
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 184) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 185) * This function should be called from the paravirt spinlock initialization
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 186) * routine.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 187) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 188) void __init __pv_init_lock_hash(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 189) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 190) int pv_hash_size = ALIGN(4 * num_possible_cpus(), PV_HE_PER_LINE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 191)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 192) if (pv_hash_size < PV_HE_MIN)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 193) pv_hash_size = PV_HE_MIN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 194)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 195) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 196) * Allocate space from bootmem which should be page-size aligned
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 197) * and hence cacheline aligned.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 198) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 199) pv_lock_hash = alloc_large_system_hash("PV qspinlock",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 200) sizeof(struct pv_hash_entry),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 201) pv_hash_size, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 202) HASH_EARLY | HASH_ZERO,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 203) &pv_lock_hash_bits, NULL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 204) pv_hash_size, pv_hash_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 205) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 206)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 207) #define for_each_hash_entry(he, offset, hash) \
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 208) for (hash &= ~(PV_HE_PER_LINE - 1), he = &pv_lock_hash[hash], offset = 0; \
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 209) offset < (1 << pv_lock_hash_bits); \
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 210) offset++, he = &pv_lock_hash[(hash + offset) & ((1 << pv_lock_hash_bits) - 1)])
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 211)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 212) static struct qspinlock **pv_hash(struct qspinlock *lock, struct pv_node *node)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 213) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 214) unsigned long offset, hash = hash_ptr(lock, pv_lock_hash_bits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 215) struct pv_hash_entry *he;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 216) int hopcnt = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 217)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 218) for_each_hash_entry(he, offset, hash) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 219) hopcnt++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 220) if (!cmpxchg(&he->lock, NULL, lock)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 221) WRITE_ONCE(he->node, node);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 222) lockevent_pv_hop(hopcnt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 223) return &he->lock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 224) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 225) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 226) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 227) * Hard assume there is a free entry for us.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 228) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 229) * This is guaranteed by ensuring every blocked lock only ever consumes
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 230) * a single entry, and since we only have 4 nesting levels per CPU
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 231) * and allocated 4*nr_possible_cpus(), this must be so.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 232) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 233) * The single entry is guaranteed by having the lock owner unhash
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 234) * before it releases.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 235) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 236) BUG();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 237) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 238)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 239) static struct pv_node *pv_unhash(struct qspinlock *lock)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 240) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 241) unsigned long offset, hash = hash_ptr(lock, pv_lock_hash_bits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 242) struct pv_hash_entry *he;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 243) struct pv_node *node;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 244)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 245) for_each_hash_entry(he, offset, hash) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 246) if (READ_ONCE(he->lock) == lock) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 247) node = READ_ONCE(he->node);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 248) WRITE_ONCE(he->lock, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 249) return node;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 250) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 251) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 252) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 253) * Hard assume we'll find an entry.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 254) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 255) * This guarantees a limited lookup time and is itself guaranteed by
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 256) * having the lock owner do the unhash -- IFF the unlock sees the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 257) * SLOW flag, there MUST be a hash entry.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 258) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 259) BUG();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 260) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 261)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 262) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 263) * Return true if when it is time to check the previous node which is not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 264) * in a running state.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 265) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 266) static inline bool
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 267) pv_wait_early(struct pv_node *prev, int loop)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 268) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 269) if ((loop & PV_PREV_CHECK_MASK) != 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 270) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 271)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 272) return READ_ONCE(prev->state) != vcpu_running;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 273) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 274)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 275) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 276) * Initialize the PV part of the mcs_spinlock node.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 277) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 278) static void pv_init_node(struct mcs_spinlock *node)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 279) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 280) struct pv_node *pn = (struct pv_node *)node;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 281)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 282) BUILD_BUG_ON(sizeof(struct pv_node) > sizeof(struct qnode));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 283)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 284) pn->cpu = smp_processor_id();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 285) pn->state = vcpu_running;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 286) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 287)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 288) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 289) * Wait for node->locked to become true, halt the vcpu after a short spin.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 290) * pv_kick_node() is used to set _Q_SLOW_VAL and fill in hash table on its
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 291) * behalf.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 292) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 293) static void pv_wait_node(struct mcs_spinlock *node, struct mcs_spinlock *prev)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 294) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 295) struct pv_node *pn = (struct pv_node *)node;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 296) struct pv_node *pp = (struct pv_node *)prev;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 297) int loop;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 298) bool wait_early;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 299)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 300) for (;;) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 301) for (wait_early = false, loop = SPIN_THRESHOLD; loop; loop--) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 302) if (READ_ONCE(node->locked))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 303) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 304) if (pv_wait_early(pp, loop)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 305) wait_early = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 306) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 307) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 308) cpu_relax();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 309) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 310)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 311) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 312) * Order pn->state vs pn->locked thusly:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 313) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 314) * [S] pn->state = vcpu_halted [S] next->locked = 1
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 315) * MB MB
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 316) * [L] pn->locked [RmW] pn->state = vcpu_hashed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 317) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 318) * Matches the cmpxchg() from pv_kick_node().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 319) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 320) smp_store_mb(pn->state, vcpu_halted);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 321)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 322) if (!READ_ONCE(node->locked)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 323) lockevent_inc(pv_wait_node);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 324) lockevent_cond_inc(pv_wait_early, wait_early);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 325) pv_wait(&pn->state, vcpu_halted);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 326) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 327)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 328) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 329) * If pv_kick_node() changed us to vcpu_hashed, retain that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 330) * value so that pv_wait_head_or_lock() knows to not also try
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 331) * to hash this lock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 332) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 333) cmpxchg(&pn->state, vcpu_halted, vcpu_running);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 334)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 335) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 336) * If the locked flag is still not set after wakeup, it is a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 337) * spurious wakeup and the vCPU should wait again. However,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 338) * there is a pretty high overhead for CPU halting and kicking.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 339) * So it is better to spin for a while in the hope that the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 340) * MCS lock will be released soon.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 341) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 342) lockevent_cond_inc(pv_spurious_wakeup,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 343) !READ_ONCE(node->locked));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 344) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 345)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 346) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 347) * By now our node->locked should be 1 and our caller will not actually
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 348) * spin-wait for it. We do however rely on our caller to do a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 349) * load-acquire for us.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 350) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 351) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 352)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 353) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 354) * Called after setting next->locked = 1 when we're the lock owner.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 355) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 356) * Instead of waking the waiters stuck in pv_wait_node() advance their state
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 357) * such that they're waiting in pv_wait_head_or_lock(), this avoids a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 358) * wake/sleep cycle.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 359) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 360) static void pv_kick_node(struct qspinlock *lock, struct mcs_spinlock *node)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 361) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 362) struct pv_node *pn = (struct pv_node *)node;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 363)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 364) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 365) * If the vCPU is indeed halted, advance its state to match that of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 366) * pv_wait_node(). If OTOH this fails, the vCPU was running and will
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 367) * observe its next->locked value and advance itself.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 368) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 369) * Matches with smp_store_mb() and cmpxchg() in pv_wait_node()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 370) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 371) * The write to next->locked in arch_mcs_spin_unlock_contended()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 372) * must be ordered before the read of pn->state in the cmpxchg()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 373) * below for the code to work correctly. To guarantee full ordering
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 374) * irrespective of the success or failure of the cmpxchg(),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 375) * a relaxed version with explicit barrier is used. The control
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 376) * dependency will order the reading of pn->state before any
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 377) * subsequent writes.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 378) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 379) smp_mb__before_atomic();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 380) if (cmpxchg_relaxed(&pn->state, vcpu_halted, vcpu_hashed)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 381) != vcpu_halted)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 382) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 383)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 384) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 385) * Put the lock into the hash table and set the _Q_SLOW_VAL.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 386) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 387) * As this is the same vCPU that will check the _Q_SLOW_VAL value and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 388) * the hash table later on at unlock time, no atomic instruction is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 389) * needed.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 390) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 391) WRITE_ONCE(lock->locked, _Q_SLOW_VAL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 392) (void)pv_hash(lock, pn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 393) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 394)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 395) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 396) * Wait for l->locked to become clear and acquire the lock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 397) * halt the vcpu after a short spin.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 398) * __pv_queued_spin_unlock() will wake us.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 399) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 400) * The current value of the lock will be returned for additional processing.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 401) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 402) static u32
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 403) pv_wait_head_or_lock(struct qspinlock *lock, struct mcs_spinlock *node)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 404) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 405) struct pv_node *pn = (struct pv_node *)node;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 406) struct qspinlock **lp = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 407) int waitcnt = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 408) int loop;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 409)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 410) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 411) * If pv_kick_node() already advanced our state, we don't need to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 412) * insert ourselves into the hash table anymore.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 413) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 414) if (READ_ONCE(pn->state) == vcpu_hashed)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 415) lp = (struct qspinlock **)1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 416)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 417) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 418) * Tracking # of slowpath locking operations
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 419) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 420) lockevent_inc(lock_slowpath);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 421)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 422) for (;; waitcnt++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 423) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 424) * Set correct vCPU state to be used by queue node wait-early
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 425) * mechanism.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 426) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 427) WRITE_ONCE(pn->state, vcpu_running);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 428)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 429) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 430) * Set the pending bit in the active lock spinning loop to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 431) * disable lock stealing before attempting to acquire the lock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 432) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 433) set_pending(lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 434) for (loop = SPIN_THRESHOLD; loop; loop--) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 435) if (trylock_clear_pending(lock))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 436) goto gotlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 437) cpu_relax();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 438) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 439) clear_pending(lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 440)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 441)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 442) if (!lp) { /* ONCE */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 443) lp = pv_hash(lock, pn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 444)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 445) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 446) * We must hash before setting _Q_SLOW_VAL, such that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 447) * when we observe _Q_SLOW_VAL in __pv_queued_spin_unlock()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 448) * we'll be sure to be able to observe our hash entry.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 449) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 450) * [S] <hash> [Rmw] l->locked == _Q_SLOW_VAL
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 451) * MB RMB
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 452) * [RmW] l->locked = _Q_SLOW_VAL [L] <unhash>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 453) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 454) * Matches the smp_rmb() in __pv_queued_spin_unlock().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 455) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 456) if (xchg(&lock->locked, _Q_SLOW_VAL) == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 457) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 458) * The lock was free and now we own the lock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 459) * Change the lock value back to _Q_LOCKED_VAL
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 460) * and unhash the table.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 461) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 462) WRITE_ONCE(lock->locked, _Q_LOCKED_VAL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 463) WRITE_ONCE(*lp, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 464) goto gotlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 465) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 466) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 467) WRITE_ONCE(pn->state, vcpu_hashed);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 468) lockevent_inc(pv_wait_head);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 469) lockevent_cond_inc(pv_wait_again, waitcnt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 470) pv_wait(&lock->locked, _Q_SLOW_VAL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 471)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 472) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 473) * Because of lock stealing, the queue head vCPU may not be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 474) * able to acquire the lock before it has to wait again.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 475) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 476) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 477)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 478) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 479) * The cmpxchg() or xchg() call before coming here provides the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 480) * acquire semantics for locking. The dummy ORing of _Q_LOCKED_VAL
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 481) * here is to indicate to the compiler that the value will always
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 482) * be nozero to enable better code optimization.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 483) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 484) gotlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 485) return (u32)(atomic_read(&lock->val) | _Q_LOCKED_VAL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 486) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 487)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 488) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 489) * PV versions of the unlock fastpath and slowpath functions to be used
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 490) * instead of queued_spin_unlock().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 491) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 492) __visible void
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 493) __pv_queued_spin_unlock_slowpath(struct qspinlock *lock, u8 locked)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 494) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 495) struct pv_node *node;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 496)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 497) if (unlikely(locked != _Q_SLOW_VAL)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 498) WARN(!debug_locks_silent,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 499) "pvqspinlock: lock 0x%lx has corrupted value 0x%x!\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 500) (unsigned long)lock, atomic_read(&lock->val));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 501) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 502) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 503)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 504) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 505) * A failed cmpxchg doesn't provide any memory-ordering guarantees,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 506) * so we need a barrier to order the read of the node data in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 507) * pv_unhash *after* we've read the lock being _Q_SLOW_VAL.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 508) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 509) * Matches the cmpxchg() in pv_wait_head_or_lock() setting _Q_SLOW_VAL.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 510) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 511) smp_rmb();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 512)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 513) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 514) * Since the above failed to release, this must be the SLOW path.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 515) * Therefore start by looking up the blocked node and unhashing it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 516) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 517) node = pv_unhash(lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 518)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 519) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 520) * Now that we have a reference to the (likely) blocked pv_node,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 521) * release the lock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 522) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 523) smp_store_release(&lock->locked, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 524)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 525) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 526) * At this point the memory pointed at by lock can be freed/reused,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 527) * however we can still use the pv_node to kick the CPU.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 528) * The other vCPU may not really be halted, but kicking an active
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 529) * vCPU is harmless other than the additional latency in completing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 530) * the unlock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 531) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 532) lockevent_inc(pv_kick_unlock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 533) pv_kick(node->cpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 534) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 535)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 536) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 537) * Include the architecture specific callee-save thunk of the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 538) * __pv_queued_spin_unlock(). This thunk is put together with
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 539) * __pv_queued_spin_unlock() to make the callee-save thunk and the real unlock
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 540) * function close to each other sharing consecutive instruction cachelines.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 541) * Alternatively, architecture specific version of __pv_queued_spin_unlock()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 542) * can be defined.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 543) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 544) #include <asm/qspinlock_paravirt.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 545)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 546) #ifndef __pv_queued_spin_unlock
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 547) __visible void __pv_queued_spin_unlock(struct qspinlock *lock)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 548) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 549) u8 locked;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 550)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 551) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 552) * We must not unlock if SLOW, because in that case we must first
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 553) * unhash. Otherwise it would be possible to have multiple @lock
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 554) * entries, which would be BAD.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 555) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 556) locked = cmpxchg_release(&lock->locked, _Q_LOCKED_VAL, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 557) if (likely(locked == _Q_LOCKED_VAL))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 558) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 559)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 560) __pv_queued_spin_unlock_slowpath(lock, locked);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 561) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 562) #endif /* __pv_queued_spin_unlock */
Orange Pi5 kernel
Deprecated Linux kernel 5.10.110 for OrangePi 5/5B/5+ boards
3 Commits
0 Branches
0 Tags