mirror of
https://github.com/AuxXxilium/linux_dsm_epyc7002.git
synced 2024-12-20 11:47:44 +07:00
726328d92a
This patch updates/fixes all spin_unlock_wait() implementations. The update is in semantics; where it previously was only a control dependency, we now upgrade to a full load-acquire to match the store-release from the spin_unlock() we waited on. This ensures that when spin_unlock_wait() returns, we're guaranteed to observe the full critical section we waited on. This fixes a number of spin_unlock_wait() users that (not unreasonably) rely on this. I also fixed a number of ticket lock versions to only wait on the current lock holder, instead of for a full unlock, as this is sufficient. Furthermore; again for ticket locks; I added an smp_rmb() in between the initial ticket load and the spin loop testing the current value because I could not convince myself the address dependency is sufficient, esp. if the loads are of different sizes. I'm more than happy to remove this smp_rmb() again if people are certain the address dependency does indeed work as expected. Note: PPC32 will be fixed independently Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: chris@zankel.net Cc: cmetcalf@mellanox.com Cc: davem@davemloft.net Cc: dhowells@redhat.com Cc: james.hogan@imgtec.com Cc: jejb@parisc-linux.org Cc: linux@armlinux.org.uk Cc: mpe@ellerman.id.au Cc: ralf@linux-mips.org Cc: realmz6@gmail.com Cc: rkuo@codeaurora.org Cc: rth@twiddle.net Cc: schwidefsky@de.ibm.com Cc: tony.luck@intel.com Cc: vgupta@synopsys.com Cc: ysato@users.sourceforge.jp Cc: linux-kernel@vger.kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org>
212 lines
5.0 KiB
C
212 lines
5.0 KiB
C
/*
|
|
* include/asm-xtensa/spinlock.h
|
|
*
|
|
* This file is subject to the terms and conditions of the GNU General Public
|
|
* License. See the file "COPYING" in the main directory of this archive
|
|
* for more details.
|
|
*
|
|
* Copyright (C) 2001 - 2005 Tensilica Inc.
|
|
*/
|
|
|
|
#ifndef _XTENSA_SPINLOCK_H
|
|
#define _XTENSA_SPINLOCK_H
|
|
|
|
#include <asm/barrier.h>
|
|
#include <asm/processor.h>
|
|
|
|
/*
|
|
* spinlock
|
|
*
|
|
* There is at most one owner of a spinlock. There are not different
|
|
* types of spinlock owners like there are for rwlocks (see below).
|
|
*
|
|
* When trying to obtain a spinlock, the function "spins" forever, or busy-
|
|
* waits, until the lock is obtained. When spinning, presumably some other
|
|
* owner will soon give up the spinlock making it available to others. Use
|
|
* the trylock functions to avoid spinning forever.
|
|
*
|
|
* possible values:
|
|
*
|
|
* 0 nobody owns the spinlock
|
|
* 1 somebody owns the spinlock
|
|
*/
|
|
|
|
#define arch_spin_is_locked(x) ((x)->slock != 0)
|
|
|
|
static inline void arch_spin_unlock_wait(arch_spinlock_t *lock)
|
|
{
|
|
smp_cond_load_acquire(&lock->slock, !VAL);
|
|
}
|
|
|
|
#define arch_spin_lock_flags(lock, flags) arch_spin_lock(lock)
|
|
|
|
static inline void arch_spin_lock(arch_spinlock_t *lock)
|
|
{
|
|
unsigned long tmp;
|
|
|
|
__asm__ __volatile__(
|
|
" movi %0, 0\n"
|
|
" wsr %0, scompare1\n"
|
|
"1: movi %0, 1\n"
|
|
" s32c1i %0, %1, 0\n"
|
|
" bnez %0, 1b\n"
|
|
: "=&a" (tmp)
|
|
: "a" (&lock->slock)
|
|
: "memory");
|
|
}
|
|
|
|
/* Returns 1 if the lock is obtained, 0 otherwise. */
|
|
|
|
static inline int arch_spin_trylock(arch_spinlock_t *lock)
|
|
{
|
|
unsigned long tmp;
|
|
|
|
__asm__ __volatile__(
|
|
" movi %0, 0\n"
|
|
" wsr %0, scompare1\n"
|
|
" movi %0, 1\n"
|
|
" s32c1i %0, %1, 0\n"
|
|
: "=&a" (tmp)
|
|
: "a" (&lock->slock)
|
|
: "memory");
|
|
|
|
return tmp == 0 ? 1 : 0;
|
|
}
|
|
|
|
static inline void arch_spin_unlock(arch_spinlock_t *lock)
|
|
{
|
|
unsigned long tmp;
|
|
|
|
__asm__ __volatile__(
|
|
" movi %0, 0\n"
|
|
" s32ri %0, %1, 0\n"
|
|
: "=&a" (tmp)
|
|
: "a" (&lock->slock)
|
|
: "memory");
|
|
}
|
|
|
|
/*
|
|
* rwlock
|
|
*
|
|
* Read-write locks are really a more flexible spinlock. They allow
|
|
* multiple readers but only one writer. Write ownership is exclusive
|
|
* (i.e., all other readers and writers are blocked from ownership while
|
|
* there is a write owner). These rwlocks are unfair to writers. Writers
|
|
* can be starved for an indefinite time by readers.
|
|
*
|
|
* possible values:
|
|
*
|
|
* 0 nobody owns the rwlock
|
|
* >0 one or more readers own the rwlock
|
|
* (the positive value is the actual number of readers)
|
|
* 0x80000000 one writer owns the rwlock, no other writers, no readers
|
|
*/
|
|
|
|
#define arch_write_can_lock(x) ((x)->lock == 0)
|
|
|
|
static inline void arch_write_lock(arch_rwlock_t *rw)
|
|
{
|
|
unsigned long tmp;
|
|
|
|
__asm__ __volatile__(
|
|
" movi %0, 0\n"
|
|
" wsr %0, scompare1\n"
|
|
"1: movi %0, 1\n"
|
|
" slli %0, %0, 31\n"
|
|
" s32c1i %0, %1, 0\n"
|
|
" bnez %0, 1b\n"
|
|
: "=&a" (tmp)
|
|
: "a" (&rw->lock)
|
|
: "memory");
|
|
}
|
|
|
|
/* Returns 1 if the lock is obtained, 0 otherwise. */
|
|
|
|
static inline int arch_write_trylock(arch_rwlock_t *rw)
|
|
{
|
|
unsigned long tmp;
|
|
|
|
__asm__ __volatile__(
|
|
" movi %0, 0\n"
|
|
" wsr %0, scompare1\n"
|
|
" movi %0, 1\n"
|
|
" slli %0, %0, 31\n"
|
|
" s32c1i %0, %1, 0\n"
|
|
: "=&a" (tmp)
|
|
: "a" (&rw->lock)
|
|
: "memory");
|
|
|
|
return tmp == 0 ? 1 : 0;
|
|
}
|
|
|
|
static inline void arch_write_unlock(arch_rwlock_t *rw)
|
|
{
|
|
unsigned long tmp;
|
|
|
|
__asm__ __volatile__(
|
|
" movi %0, 0\n"
|
|
" s32ri %0, %1, 0\n"
|
|
: "=&a" (tmp)
|
|
: "a" (&rw->lock)
|
|
: "memory");
|
|
}
|
|
|
|
static inline void arch_read_lock(arch_rwlock_t *rw)
|
|
{
|
|
unsigned long tmp;
|
|
unsigned long result;
|
|
|
|
__asm__ __volatile__(
|
|
"1: l32i %1, %2, 0\n"
|
|
" bltz %1, 1b\n"
|
|
" wsr %1, scompare1\n"
|
|
" addi %0, %1, 1\n"
|
|
" s32c1i %0, %2, 0\n"
|
|
" bne %0, %1, 1b\n"
|
|
: "=&a" (result), "=&a" (tmp)
|
|
: "a" (&rw->lock)
|
|
: "memory");
|
|
}
|
|
|
|
/* Returns 1 if the lock is obtained, 0 otherwise. */
|
|
|
|
static inline int arch_read_trylock(arch_rwlock_t *rw)
|
|
{
|
|
unsigned long result;
|
|
unsigned long tmp;
|
|
|
|
__asm__ __volatile__(
|
|
" l32i %1, %2, 0\n"
|
|
" addi %0, %1, 1\n"
|
|
" bltz %0, 1f\n"
|
|
" wsr %1, scompare1\n"
|
|
" s32c1i %0, %2, 0\n"
|
|
" sub %0, %0, %1\n"
|
|
"1:\n"
|
|
: "=&a" (result), "=&a" (tmp)
|
|
: "a" (&rw->lock)
|
|
: "memory");
|
|
|
|
return result == 0;
|
|
}
|
|
|
|
static inline void arch_read_unlock(arch_rwlock_t *rw)
|
|
{
|
|
unsigned long tmp1, tmp2;
|
|
|
|
__asm__ __volatile__(
|
|
"1: l32i %1, %2, 0\n"
|
|
" addi %0, %1, -1\n"
|
|
" wsr %1, scompare1\n"
|
|
" s32c1i %0, %2, 0\n"
|
|
" bne %0, %1, 1b\n"
|
|
: "=&a" (tmp1), "=&a" (tmp2)
|
|
: "a" (&rw->lock)
|
|
: "memory");
|
|
}
|
|
|
|
#define arch_read_lock_flags(lock, flags) arch_read_lock(lock)
|
|
#define arch_write_lock_flags(lock, flags) arch_write_lock(lock)
|
|
|
|
#endif /* _XTENSA_SPINLOCK_H */
|