[deliverable/linux.git] / arch / x86 / include / asm / spinlock.h

#ifndef _ASM_X86_SPINLOCK_H
#define _ASM_X86_SPINLOCK_H

#include <linux/jump_label.h>
#include <linux/atomic.h>
#include <asm/page.h>
#include <asm/processor.h>
#include <linux/compiler.h>
#include <asm/paravirt.h>
#include <asm/bitops.h>

/*
 * Your basic SMP spinlocks, allowing only a single CPU anywhere
 *
 * Simple spin lock operations.  There are two variants, one clears IRQ's
 * on the local processor, one does not.
 *
 * These are fair FIFO ticket locks, which support up to 2^16 CPUs.
 *
 * (the type definitions are in asm/spinlock_types.h)
 */

#ifdef CONFIG_X86_32
# define LOCK_PTR_REG "a"
#else
# define LOCK_PTR_REG "D"
#endif

#if defined(CONFIG_X86_32) && (defined(CONFIG_X86_PPRO_FENCE))
/*
 * On PPro SMP, we use a locked operation to unlock
 * (PPro errata 66, 92)
 */
# define UNLOCK_LOCK_PREFIX LOCK_PREFIX
#else
# define UNLOCK_LOCK_PREFIX
#endif

/* How long a lock should spin before we consider blocking */
#define SPIN_THRESHOLD	(1 << 15)

extern struct static_key paravirt_ticketlocks_enabled;
static __always_inline bool static_key_false(struct static_key *key);

#ifdef CONFIG_QUEUED_SPINLOCKS
#include <asm/qspinlock.h>
#else

#ifdef CONFIG_PARAVIRT_SPINLOCKS

static inline void __ticket_enter_slowpath(arch_spinlock_t *lock)
{
	set_bit(0, (volatile unsigned long *)&lock->tickets.head);
}

#else  /* !CONFIG_PARAVIRT_SPINLOCKS */
static __always_inline void __ticket_lock_spinning(arch_spinlock_t *lock,
							__ticket_t ticket)
{
}
static inline void __ticket_unlock_kick(arch_spinlock_t *lock,
							__ticket_t ticket)
{
}

#endif /* CONFIG_PARAVIRT_SPINLOCKS */
static inline int  __tickets_equal(__ticket_t one, __ticket_t two)
{
	return !((one ^ two) & ~TICKET_SLOWPATH_FLAG);
}

static inline void __ticket_check_and_clear_slowpath(arch_spinlock_t *lock,
							__ticket_t head)
{
	if (head & TICKET_SLOWPATH_FLAG) {
		arch_spinlock_t old, new;

		old.tickets.head = head;
		new.tickets.head = head & ~TICKET_SLOWPATH_FLAG;
		old.tickets.tail = new.tickets.head + TICKET_LOCK_INC;
		new.tickets.tail = old.tickets.tail;

		/* try to clear slowpath flag when there are no contenders */
		cmpxchg(&lock->head_tail, old.head_tail, new.head_tail);
	}
}

static __always_inline int arch_spin_value_unlocked(arch_spinlock_t lock)
{
	return __tickets_equal(lock.tickets.head, lock.tickets.tail);
}

/*
 * Ticket locks are conceptually two parts, one indicating the current head of
 * the queue, and the other indicating the current tail. The lock is acquired
 * by atomically noting the tail and incrementing it by one (thus adding
 * ourself to the queue and noting our position), then waiting until the head
 * becomes equal to the the initial value of the tail.
 *
 * We use an xadd covering *both* parts of the lock, to increment the tail and
 * also load the position of the head, which takes care of memory ordering
 * issues and should be optimal for the uncontended case. Note the tail must be
 * in the high part, because a wide xadd increment of the low part would carry
 * up and contaminate the high part.
 */
static __always_inline void arch_spin_lock(arch_spinlock_t *lock)
{
	register struct __raw_tickets inc = { .tail = TICKET_LOCK_INC };

	inc = xadd(&lock->tickets, inc);
	if (likely(inc.head == inc.tail))
		goto out;

	for (;;) {
		unsigned count = SPIN_THRESHOLD;

		do {
			inc.head = READ_ONCE(lock->tickets.head);
			if (__tickets_equal(inc.head, inc.tail))
				goto clear_slowpath;
			cpu_relax();
		} while (--count);
		__ticket_lock_spinning(lock, inc.tail);
	}
clear_slowpath:
	__ticket_check_and_clear_slowpath(lock, inc.head);
out:
	barrier();	/* make sure nothing creeps before the lock is taken */
}

static __always_inline int arch_spin_trylock(arch_spinlock_t *lock)
{
	arch_spinlock_t old, new;

	old.tickets = READ_ONCE(lock->tickets);
	if (!__tickets_equal(old.tickets.head, old.tickets.tail))
		return 0;

	new.head_tail = old.head_tail + (TICKET_LOCK_INC << TICKET_SHIFT);
	new.head_tail &= ~TICKET_SLOWPATH_FLAG;

	/* cmpxchg is a full barrier, so nothing can move before it */
	return cmpxchg(&lock->head_tail, old.head_tail, new.head_tail) == old.head_tail;
}

static __always_inline void arch_spin_unlock(arch_spinlock_t *lock)
{
	if (TICKET_SLOWPATH_FLAG &&
		static_key_false(&paravirt_ticketlocks_enabled)) {
		__ticket_t head;

		BUILD_BUG_ON(((__ticket_t)NR_CPUS) != NR_CPUS);

		head = xadd(&lock->tickets.head, TICKET_LOCK_INC);

		if (unlikely(head & TICKET_SLOWPATH_FLAG)) {
			head &= ~TICKET_SLOWPATH_FLAG;
			__ticket_unlock_kick(lock, (head + TICKET_LOCK_INC));
		}
	} else
		__add(&lock->tickets.head, TICKET_LOCK_INC, UNLOCK_LOCK_PREFIX);
}

static inline int arch_spin_is_locked(arch_spinlock_t *lock)
{
	struct __raw_tickets tmp = READ_ONCE(lock->tickets);

	return !__tickets_equal(tmp.tail, tmp.head);
}

static inline int arch_spin_is_contended(arch_spinlock_t *lock)
{
	struct __raw_tickets tmp = READ_ONCE(lock->tickets);

	tmp.head &= ~TICKET_SLOWPATH_FLAG;
	return (__ticket_t)(tmp.tail - tmp.head) > TICKET_LOCK_INC;
}
#define arch_spin_is_contended	arch_spin_is_contended

static __always_inline void arch_spin_lock_flags(arch_spinlock_t *lock,
						  unsigned long flags)
{
	arch_spin_lock(lock);
}

static inline void arch_spin_unlock_wait(arch_spinlock_t *lock)
{
	__ticket_t head = READ_ONCE(lock->tickets.head);

	for (;;) {
		struct __raw_tickets tmp = READ_ONCE(lock->tickets);
		/*
		 * We need to check "unlocked" in a loop, tmp.head == head
		 * can be false positive because of overflow.
		 */
		if (__tickets_equal(tmp.head, tmp.tail) ||
				!__tickets_equal(tmp.head, head))
			break;

		cpu_relax();
	}
}
#endif /* CONFIG_QUEUED_SPINLOCKS */

/*
 * Read-write spinlocks, allowing multiple readers
 * but only one writer.
 *
 * NOTE! it is quite common to have readers in interrupts
 * but no interrupt writers. For those circumstances we
 * can "mix" irq-safe locks - any writer needs to get a
 * irq-safe write-lock, but readers can get non-irqsafe
 * read-locks.
 *
 * On x86, we implement read-write locks using the generic qrwlock with
 * x86 specific optimization.
 */

#include <asm/qrwlock.h>

#define arch_read_lock_flags(lock, flags) arch_read_lock(lock)
#define arch_write_lock_flags(lock, flags) arch_write_lock(lock)

#define arch_spin_relax(lock)	cpu_relax()
#define arch_read_relax(lock)	cpu_relax()
#define arch_write_relax(lock)	cpu_relax()

#endif /* _ASM_X86_SPINLOCK_H */
Commit	Line	Data
	1	#ifndef _ASM_X86_SPINLOCK_H
	2	#define _ASM_X86_SPINLOCK_H
	3
	4	#include <linux/jump_label.h>
	5	#include <linux/atomic.h>
	6	#include <asm/page.h>
	7	#include <asm/processor.h>
	8	#include <linux/compiler.h>
	9	#include <asm/paravirt.h>
	10	#include <asm/bitops.h>
	11
	12	/*
	13	* Your basic SMP spinlocks, allowing only a single CPU anywhere
	14	*
	15	* Simple spin lock operations. There are two variants, one clears IRQ's
	16	* on the local processor, one does not.
	17	*
	18	* These are fair FIFO ticket locks, which support up to 2^16 CPUs.
	19	*
	20	* (the type definitions are in asm/spinlock_types.h)
	21	*/
	22
	23	#ifdef CONFIG_X86_32
	24	# define LOCK_PTR_REG "a"
	25	#else
	26	# define LOCK_PTR_REG "D"
	27	#endif
	28
	29	#if defined(CONFIG_X86_32) && (defined(CONFIG_X86_PPRO_FENCE))
	30	/*
	31	* On PPro SMP, we use a locked operation to unlock
	32	* (PPro errata 66, 92)
	33	*/
	34	# define UNLOCK_LOCK_PREFIX LOCK_PREFIX
	35	#else
	36	# define UNLOCK_LOCK_PREFIX
	37	#endif
	38
	39	/* How long a lock should spin before we consider blocking */
	40	#define SPIN_THRESHOLD (1 << 15)
	41
	42	extern struct static_key paravirt_ticketlocks_enabled;
	43	static __always_inline bool static_key_false(struct static_key *key);
	44
	45	#ifdef CONFIG_QUEUED_SPINLOCKS
	46	#include <asm/qspinlock.h>
	47	#else
	48
	49	#ifdef CONFIG_PARAVIRT_SPINLOCKS
	50
	51	static inline void __ticket_enter_slowpath(arch_spinlock_t *lock)
	52	{
	53	set_bit(0, (volatile unsigned long *)&lock->tickets.head);
	54	}
	55
	56	#else /* !CONFIG_PARAVIRT_SPINLOCKS */
	57	static __always_inline void __ticket_lock_spinning(arch_spinlock_t *lock,
	58	__ticket_t ticket)
	59	{
	60	}
	61	static inline void __ticket_unlock_kick(arch_spinlock_t *lock,
	62	__ticket_t ticket)
	63	{
	64	}
	65
	66	#endif /* CONFIG_PARAVIRT_SPINLOCKS */
	67	static inline int __tickets_equal(__ticket_t one, __ticket_t two)
	68	{
	69	return !((one ^ two) & ~TICKET_SLOWPATH_FLAG);
	70	}
	71
	72	static inline void __ticket_check_and_clear_slowpath(arch_spinlock_t *lock,
	73	__ticket_t head)
	74	{
	75	if (head & TICKET_SLOWPATH_FLAG) {
	76	arch_spinlock_t old, new;
	77
	78	old.tickets.head = head;
	79	new.tickets.head = head & ~TICKET_SLOWPATH_FLAG;
	80	old.tickets.tail = new.tickets.head + TICKET_LOCK_INC;
	81	new.tickets.tail = old.tickets.tail;
	82
	83	/* try to clear slowpath flag when there are no contenders */
	84	cmpxchg(&lock->head_tail, old.head_tail, new.head_tail);
	85	}
	86	}
	87
	88	static __always_inline int arch_spin_value_unlocked(arch_spinlock_t lock)
	89	{
	90	return __tickets_equal(lock.tickets.head, lock.tickets.tail);
	91	}
	92
	93	/*
	94	* Ticket locks are conceptually two parts, one indicating the current head of
	95	* the queue, and the other indicating the current tail. The lock is acquired
	96	* by atomically noting the tail and incrementing it by one (thus adding
	97	* ourself to the queue and noting our position), then waiting until the head
	98	* becomes equal to the the initial value of the tail.
	99	*
	100	* We use an xadd covering both parts of the lock, to increment the tail and
	101	* also load the position of the head, which takes care of memory ordering
	102	* issues and should be optimal for the uncontended case. Note the tail must be
	103	* in the high part, because a wide xadd increment of the low part would carry
	104	* up and contaminate the high part.
	105	*/
	106	static __always_inline void arch_spin_lock(arch_spinlock_t *lock)
	107	{
	108	register struct __raw_tickets inc = { .tail = TICKET_LOCK_INC };
	109
	110	inc = xadd(&lock->tickets, inc);
	111	if (likely(inc.head == inc.tail))
	112	goto out;
	113
	114	for (;;) {
	115	unsigned count = SPIN_THRESHOLD;
	116
	117	do {
	118	inc.head = READ_ONCE(lock->tickets.head);
	119	if (__tickets_equal(inc.head, inc.tail))
	120	goto clear_slowpath;
	121	cpu_relax();
	122	} while (--count);
	123	__ticket_lock_spinning(lock, inc.tail);
	124	}
	125	clear_slowpath:
	126	__ticket_check_and_clear_slowpath(lock, inc.head);
	127	out:
	128	barrier(); /* make sure nothing creeps before the lock is taken */
	129	}
	130
	131	static __always_inline int arch_spin_trylock(arch_spinlock_t *lock)
	132	{
	133	arch_spinlock_t old, new;
	134
	135	old.tickets = READ_ONCE(lock->tickets);
	136	if (!__tickets_equal(old.tickets.head, old.tickets.tail))
	137	return 0;
	138
	139	new.head_tail = old.head_tail + (TICKET_LOCK_INC << TICKET_SHIFT);
	140	new.head_tail &= ~TICKET_SLOWPATH_FLAG;
	141
	142	/* cmpxchg is a full barrier, so nothing can move before it */
	143	return cmpxchg(&lock->head_tail, old.head_tail, new.head_tail) == old.head_tail;
	144	}
	145
	146	static __always_inline void arch_spin_unlock(arch_spinlock_t *lock)
	147	{
	148	if (TICKET_SLOWPATH_FLAG &&
	149	static_key_false(&paravirt_ticketlocks_enabled)) {
	150	__ticket_t head;
	151
	152	BUILD_BUG_ON(((__ticket_t)NR_CPUS) != NR_CPUS);
	153
	154	head = xadd(&lock->tickets.head, TICKET_LOCK_INC);
	155
	156	if (unlikely(head & TICKET_SLOWPATH_FLAG)) {
	157	head &= ~TICKET_SLOWPATH_FLAG;
	158	__ticket_unlock_kick(lock, (head + TICKET_LOCK_INC));
	159	}
	160	} else
	161	__add(&lock->tickets.head, TICKET_LOCK_INC, UNLOCK_LOCK_PREFIX);
	162	}
	163
	164	static inline int arch_spin_is_locked(arch_spinlock_t *lock)
	165	{
	166	struct __raw_tickets tmp = READ_ONCE(lock->tickets);
	167
	168	return !__tickets_equal(tmp.tail, tmp.head);
	169	}
	170
	171	static inline int arch_spin_is_contended(arch_spinlock_t *lock)
	172	{
	173	struct __raw_tickets tmp = READ_ONCE(lock->tickets);
	174
	175	tmp.head &= ~TICKET_SLOWPATH_FLAG;
	176	return (__ticket_t)(tmp.tail - tmp.head) > TICKET_LOCK_INC;
	177	}
	178	#define arch_spin_is_contended arch_spin_is_contended
	179
	180	static __always_inline void arch_spin_lock_flags(arch_spinlock_t *lock,
	181	unsigned long flags)
	182	{
	183	arch_spin_lock(lock);
	184	}
	185
	186	static inline void arch_spin_unlock_wait(arch_spinlock_t *lock)
	187	{
	188	__ticket_t head = READ_ONCE(lock->tickets.head);
	189
	190	for (;;) {
	191	struct __raw_tickets tmp = READ_ONCE(lock->tickets);
	192	/*
	193	* We need to check "unlocked" in a loop, tmp.head == head
	194	* can be false positive because of overflow.
	195	*/
	196	if (__tickets_equal(tmp.head, tmp.tail) \|\|
	197	!__tickets_equal(tmp.head, head))
	198	break;
	199
	200	cpu_relax();
	201	}
	202	}
	203	#endif /* CONFIG_QUEUED_SPINLOCKS */
	204
	205	/*
	206	* Read-write spinlocks, allowing multiple readers
	207	* but only one writer.
	208	*
	209	* NOTE! it is quite common to have readers in interrupts
	210	* but no interrupt writers. For those circumstances we
	211	* can "mix" irq-safe locks - any writer needs to get a
	212	* irq-safe write-lock, but readers can get non-irqsafe
	213	* read-locks.
	214	*
	215	* On x86, we implement read-write locks using the generic qrwlock with
	216	* x86 specific optimization.
	217	*/
	218
	219	#include <asm/qrwlock.h>
	220
	221	#define arch_read_lock_flags(lock, flags) arch_read_lock(lock)
	222	#define arch_write_lock_flags(lock, flags) arch_write_lock(lock)
	223
	224	#define arch_spin_relax(lock) cpu_relax()
	225	#define arch_read_relax(lock) cpu_relax()
	226	#define arch_write_relax(lock) cpu_relax()
	227
	228	#endif /* _ASM_X86_SPINLOCK_H */