mirror of
https://github.com/torvalds/linux.git
synced 2024-11-17 01:22:07 +00:00
4f07118f65
It appears that a memory barrier soon after a mispredicted branch, not just in the delay slot, can cause the hang condition of this cpu errata. So move them out-of-line, and explicitly put them into a "branch always, predict taken" delay slot which should fully kill this problem. Signed-off-by: David S. Miller <davem@davemloft.net>
342 lines
8.0 KiB
C
342 lines
8.0 KiB
C
/* spinlock.h: 64-bit Sparc spinlock support.
|
|
*
|
|
* Copyright (C) 1997 David S. Miller (davem@caip.rutgers.edu)
|
|
*/
|
|
|
|
#ifndef __SPARC64_SPINLOCK_H
|
|
#define __SPARC64_SPINLOCK_H
|
|
|
|
#include <linux/config.h>
|
|
#include <linux/threads.h> /* For NR_CPUS */
|
|
|
|
#ifndef __ASSEMBLY__
|
|
|
|
/* To get debugging spinlocks which detect and catch
|
|
* deadlock situations, set CONFIG_DEBUG_SPINLOCK
|
|
* and rebuild your kernel.
|
|
*/
|
|
|
|
/* All of these locking primitives are expected to work properly
|
|
* even in an RMO memory model, which currently is what the kernel
|
|
* runs in.
|
|
*
|
|
* There is another issue. Because we play games to save cycles
|
|
* in the non-contention case, we need to be extra careful about
|
|
* branch targets into the "spinning" code. They live in their
|
|
* own section, but the newer V9 branches have a shorter range
|
|
* than the traditional 32-bit sparc branch variants. The rule
|
|
* is that the branches that go into and out of the spinner sections
|
|
* must be pre-V9 branches.
|
|
*/
|
|
|
|
#ifndef CONFIG_DEBUG_SPINLOCK
|
|
|
|
typedef struct {
|
|
volatile unsigned char lock;
|
|
#ifdef CONFIG_PREEMPT
|
|
unsigned int break_lock;
|
|
#endif
|
|
} spinlock_t;
|
|
#define SPIN_LOCK_UNLOCKED (spinlock_t) {0,}
|
|
|
|
#define spin_lock_init(lp) do { *(lp)= SPIN_LOCK_UNLOCKED; } while(0)
|
|
#define spin_is_locked(lp) ((lp)->lock != 0)
|
|
|
|
#define spin_unlock_wait(lp) \
|
|
do { rmb(); \
|
|
} while((lp)->lock)
|
|
|
|
static inline void _raw_spin_lock(spinlock_t *lock)
|
|
{
|
|
unsigned long tmp;
|
|
|
|
__asm__ __volatile__(
|
|
"1: ldstub [%1], %0\n"
|
|
" membar #StoreLoad | #StoreStore\n"
|
|
" brnz,pn %0, 2f\n"
|
|
" nop\n"
|
|
" .subsection 2\n"
|
|
"2: ldub [%1], %0\n"
|
|
" membar #LoadLoad\n"
|
|
" brnz,pt %0, 2b\n"
|
|
" nop\n"
|
|
" ba,a,pt %%xcc, 1b\n"
|
|
" .previous"
|
|
: "=&r" (tmp)
|
|
: "r" (lock)
|
|
: "memory");
|
|
}
|
|
|
|
static inline int _raw_spin_trylock(spinlock_t *lock)
|
|
{
|
|
unsigned long result;
|
|
|
|
__asm__ __volatile__(
|
|
" ldstub [%1], %0\n"
|
|
" membar #StoreLoad | #StoreStore"
|
|
: "=r" (result)
|
|
: "r" (lock)
|
|
: "memory");
|
|
|
|
return (result == 0UL);
|
|
}
|
|
|
|
static inline void _raw_spin_unlock(spinlock_t *lock)
|
|
{
|
|
__asm__ __volatile__(
|
|
" membar #StoreStore | #LoadStore\n"
|
|
" stb %%g0, [%0]"
|
|
: /* No outputs */
|
|
: "r" (lock)
|
|
: "memory");
|
|
}
|
|
|
|
static inline void _raw_spin_lock_flags(spinlock_t *lock, unsigned long flags)
|
|
{
|
|
unsigned long tmp1, tmp2;
|
|
|
|
__asm__ __volatile__(
|
|
"1: ldstub [%2], %0\n"
|
|
" membar #StoreLoad | #StoreStore\n"
|
|
" brnz,pn %0, 2f\n"
|
|
" nop\n"
|
|
" .subsection 2\n"
|
|
"2: rdpr %%pil, %1\n"
|
|
" wrpr %3, %%pil\n"
|
|
"3: ldub [%2], %0\n"
|
|
" membar #LoadLoad\n"
|
|
" brnz,pt %0, 3b\n"
|
|
" nop\n"
|
|
" ba,pt %%xcc, 1b\n"
|
|
" wrpr %1, %%pil\n"
|
|
" .previous"
|
|
: "=&r" (tmp1), "=&r" (tmp2)
|
|
: "r"(lock), "r"(flags)
|
|
: "memory");
|
|
}
|
|
|
|
#else /* !(CONFIG_DEBUG_SPINLOCK) */
|
|
|
|
typedef struct {
|
|
volatile unsigned char lock;
|
|
unsigned int owner_pc, owner_cpu;
|
|
#ifdef CONFIG_PREEMPT
|
|
unsigned int break_lock;
|
|
#endif
|
|
} spinlock_t;
|
|
#define SPIN_LOCK_UNLOCKED (spinlock_t) { 0, 0, 0xff }
|
|
#define spin_lock_init(lp) do { *(lp)= SPIN_LOCK_UNLOCKED; } while(0)
|
|
#define spin_is_locked(__lock) ((__lock)->lock != 0)
|
|
#define spin_unlock_wait(__lock) \
|
|
do { \
|
|
rmb(); \
|
|
} while((__lock)->lock)
|
|
|
|
extern void _do_spin_lock(spinlock_t *lock, char *str, unsigned long caller);
|
|
extern void _do_spin_unlock(spinlock_t *lock);
|
|
extern int _do_spin_trylock(spinlock_t *lock, unsigned long caller);
|
|
|
|
#define _raw_spin_trylock(lp) \
|
|
_do_spin_trylock(lp, (unsigned long) __builtin_return_address(0))
|
|
#define _raw_spin_lock(lock) \
|
|
_do_spin_lock(lock, "spin_lock", \
|
|
(unsigned long) __builtin_return_address(0))
|
|
#define _raw_spin_unlock(lock) _do_spin_unlock(lock)
|
|
#define _raw_spin_lock_flags(lock, flags) _raw_spin_lock(lock)
|
|
|
|
#endif /* CONFIG_DEBUG_SPINLOCK */
|
|
|
|
/* Multi-reader locks, these are much saner than the 32-bit Sparc ones... */
|
|
|
|
#ifndef CONFIG_DEBUG_SPINLOCK
|
|
|
|
typedef struct {
|
|
volatile unsigned int lock;
|
|
#ifdef CONFIG_PREEMPT
|
|
unsigned int break_lock;
|
|
#endif
|
|
} rwlock_t;
|
|
#define RW_LOCK_UNLOCKED (rwlock_t) {0,}
|
|
#define rwlock_init(lp) do { *(lp) = RW_LOCK_UNLOCKED; } while(0)
|
|
|
|
static void inline __read_lock(rwlock_t *lock)
|
|
{
|
|
unsigned long tmp1, tmp2;
|
|
|
|
__asm__ __volatile__ (
|
|
"1: ldsw [%2], %0\n"
|
|
" brlz,pn %0, 2f\n"
|
|
"4: add %0, 1, %1\n"
|
|
" cas [%2], %0, %1\n"
|
|
" cmp %0, %1\n"
|
|
" membar #StoreLoad | #StoreStore\n"
|
|
" bne,pn %%icc, 1b\n"
|
|
" nop\n"
|
|
" .subsection 2\n"
|
|
"2: ldsw [%2], %0\n"
|
|
" membar #LoadLoad\n"
|
|
" brlz,pt %0, 2b\n"
|
|
" nop\n"
|
|
" ba,a,pt %%xcc, 4b\n"
|
|
" .previous"
|
|
: "=&r" (tmp1), "=&r" (tmp2)
|
|
: "r" (lock)
|
|
: "memory");
|
|
}
|
|
|
|
static void inline __read_unlock(rwlock_t *lock)
|
|
{
|
|
unsigned long tmp1, tmp2;
|
|
|
|
__asm__ __volatile__(
|
|
" membar #StoreLoad | #LoadLoad\n"
|
|
"1: lduw [%2], %0\n"
|
|
" sub %0, 1, %1\n"
|
|
" cas [%2], %0, %1\n"
|
|
" cmp %0, %1\n"
|
|
" bne,pn %%xcc, 1b\n"
|
|
" nop"
|
|
: "=&r" (tmp1), "=&r" (tmp2)
|
|
: "r" (lock)
|
|
: "memory");
|
|
}
|
|
|
|
static void inline __write_lock(rwlock_t *lock)
|
|
{
|
|
unsigned long mask, tmp1, tmp2;
|
|
|
|
mask = 0x80000000UL;
|
|
|
|
__asm__ __volatile__(
|
|
"1: lduw [%2], %0\n"
|
|
" brnz,pn %0, 2f\n"
|
|
"4: or %0, %3, %1\n"
|
|
" cas [%2], %0, %1\n"
|
|
" cmp %0, %1\n"
|
|
" membar #StoreLoad | #StoreStore\n"
|
|
" bne,pn %%icc, 1b\n"
|
|
" nop\n"
|
|
" .subsection 2\n"
|
|
"2: lduw [%2], %0\n"
|
|
" membar #LoadLoad\n"
|
|
" brnz,pt %0, 2b\n"
|
|
" nop\n"
|
|
" ba,a,pt %%xcc, 4b\n"
|
|
" .previous"
|
|
: "=&r" (tmp1), "=&r" (tmp2)
|
|
: "r" (lock), "r" (mask)
|
|
: "memory");
|
|
}
|
|
|
|
static void inline __write_unlock(rwlock_t *lock)
|
|
{
|
|
__asm__ __volatile__(
|
|
" membar #LoadStore | #StoreStore\n"
|
|
" stw %%g0, [%0]"
|
|
: /* no outputs */
|
|
: "r" (lock)
|
|
: "memory");
|
|
}
|
|
|
|
static int inline __write_trylock(rwlock_t *lock)
|
|
{
|
|
unsigned long mask, tmp1, tmp2, result;
|
|
|
|
mask = 0x80000000UL;
|
|
|
|
__asm__ __volatile__(
|
|
" mov 0, %2\n"
|
|
"1: lduw [%3], %0\n"
|
|
" brnz,pn %0, 2f\n"
|
|
" or %0, %4, %1\n"
|
|
" cas [%3], %0, %1\n"
|
|
" cmp %0, %1\n"
|
|
" membar #StoreLoad | #StoreStore\n"
|
|
" bne,pn %%icc, 1b\n"
|
|
" nop\n"
|
|
" mov 1, %2\n"
|
|
"2:"
|
|
: "=&r" (tmp1), "=&r" (tmp2), "=&r" (result)
|
|
: "r" (lock), "r" (mask)
|
|
: "memory");
|
|
|
|
return result;
|
|
}
|
|
|
|
#define _raw_read_lock(p) __read_lock(p)
|
|
#define _raw_read_unlock(p) __read_unlock(p)
|
|
#define _raw_write_lock(p) __write_lock(p)
|
|
#define _raw_write_unlock(p) __write_unlock(p)
|
|
#define _raw_write_trylock(p) __write_trylock(p)
|
|
|
|
#else /* !(CONFIG_DEBUG_SPINLOCK) */
|
|
|
|
typedef struct {
|
|
volatile unsigned long lock;
|
|
unsigned int writer_pc, writer_cpu;
|
|
unsigned int reader_pc[NR_CPUS];
|
|
#ifdef CONFIG_PREEMPT
|
|
unsigned int break_lock;
|
|
#endif
|
|
} rwlock_t;
|
|
#define RW_LOCK_UNLOCKED (rwlock_t) { 0, 0, 0xff, { } }
|
|
#define rwlock_init(lp) do { *(lp) = RW_LOCK_UNLOCKED; } while(0)
|
|
|
|
extern void _do_read_lock(rwlock_t *rw, char *str, unsigned long caller);
|
|
extern void _do_read_unlock(rwlock_t *rw, char *str, unsigned long caller);
|
|
extern void _do_write_lock(rwlock_t *rw, char *str, unsigned long caller);
|
|
extern void _do_write_unlock(rwlock_t *rw, unsigned long caller);
|
|
extern int _do_write_trylock(rwlock_t *rw, char *str, unsigned long caller);
|
|
|
|
#define _raw_read_lock(lock) \
|
|
do { unsigned long flags; \
|
|
local_irq_save(flags); \
|
|
_do_read_lock(lock, "read_lock", \
|
|
(unsigned long) __builtin_return_address(0)); \
|
|
local_irq_restore(flags); \
|
|
} while(0)
|
|
|
|
#define _raw_read_unlock(lock) \
|
|
do { unsigned long flags; \
|
|
local_irq_save(flags); \
|
|
_do_read_unlock(lock, "read_unlock", \
|
|
(unsigned long) __builtin_return_address(0)); \
|
|
local_irq_restore(flags); \
|
|
} while(0)
|
|
|
|
#define _raw_write_lock(lock) \
|
|
do { unsigned long flags; \
|
|
local_irq_save(flags); \
|
|
_do_write_lock(lock, "write_lock", \
|
|
(unsigned long) __builtin_return_address(0)); \
|
|
local_irq_restore(flags); \
|
|
} while(0)
|
|
|
|
#define _raw_write_unlock(lock) \
|
|
do { unsigned long flags; \
|
|
local_irq_save(flags); \
|
|
_do_write_unlock(lock, \
|
|
(unsigned long) __builtin_return_address(0)); \
|
|
local_irq_restore(flags); \
|
|
} while(0)
|
|
|
|
#define _raw_write_trylock(lock) \
|
|
({ unsigned long flags; \
|
|
int val; \
|
|
local_irq_save(flags); \
|
|
val = _do_write_trylock(lock, "write_trylock", \
|
|
(unsigned long) __builtin_return_address(0)); \
|
|
local_irq_restore(flags); \
|
|
val; \
|
|
})
|
|
|
|
#endif /* CONFIG_DEBUG_SPINLOCK */
|
|
|
|
#define _raw_read_trylock(lock) generic_raw_read_trylock(lock)
|
|
#define read_can_lock(rw) (!((rw)->lock & 0x80000000UL))
|
|
#define write_can_lock(rw) (!(rw)->lock)
|
|
|
|
#endif /* !(__ASSEMBLY__) */
|
|
|
|
#endif /* !(__SPARC64_SPINLOCK_H) */
|