aha/arch/sparc64/lib/dec_and_lock.S
David S. Miller b445e26cbf [SPARC64]: Avoid membar instructions in delay slots.
In particular, avoid membar instructions in the delay
slot of a jmpl instruction.

UltraSPARC-I, II, IIi, and IIe have a bug, documented in
the UltraSPARC-IIi User's Manual, Appendix K, Erratum 51

The long and short of it is that if the IMU unit misses
on a branch or jmpl, and there is a store buffer synchronizing
membar in the delay slot, the chip can stop fetching instructions.

If interrupts are enabled or some other trap is enabled, the
chip will unwedge itself, but performance will suffer.

We already had a workaround for this bug in a few spots, but
it's better to have the entire tree sanitized for this rule.

Signed-off-by: David S. Miller <davem@davemloft.net>
2005-06-27 15:42:04 -07:00

80 lines
1.5 KiB
ArmAsm

/* $Id: dec_and_lock.S,v 1.5 2001/11/18 00:12:56 davem Exp $
* dec_and_lock.S: Sparc64 version of "atomic_dec_and_lock()"
* using cas and ldstub instructions.
*
* Copyright (C) 2000 David S. Miller (davem@redhat.com)
*/
#include <linux/config.h>
#include <asm/thread_info.h>
.text
.align 64
/* CAS basically works like this:
*
* void CAS(MEM, REG1, REG2)
* {
* START_ATOMIC();
* if (*(MEM) == REG1) {
* TMP = *(MEM);
* *(MEM) = REG2;
* REG2 = TMP;
* } else
* REG2 = *(MEM);
* END_ATOMIC();
* }
*/
.globl _atomic_dec_and_lock
_atomic_dec_and_lock: /* %o0 = counter, %o1 = lock */
loop1: lduw [%o0], %g2
subcc %g2, 1, %g7
be,pn %icc, start_to_zero
nop
nzero: cas [%o0], %g2, %g7
cmp %g2, %g7
bne,pn %icc, loop1
mov 0, %g1
out:
membar #StoreLoad | #StoreStore
retl
mov %g1, %o0
start_to_zero:
#ifdef CONFIG_PREEMPT
ldsw [%g6 + TI_PRE_COUNT], %g3
add %g3, 1, %g3
stw %g3, [%g6 + TI_PRE_COUNT]
#endif
to_zero:
ldstub [%o1], %g3
membar #StoreLoad | #StoreStore
brnz,pn %g3, spin_on_lock
nop
loop2: cas [%o0], %g2, %g7 /* ASSERT(g7 == 0) */
cmp %g2, %g7
be,pt %icc, out
mov 1, %g1
lduw [%o0], %g2
subcc %g2, 1, %g7
be,pn %icc, loop2
nop
membar #StoreStore | #LoadStore
stb %g0, [%o1]
#ifdef CONFIG_PREEMPT
ldsw [%g6 + TI_PRE_COUNT], %g3
sub %g3, 1, %g3
stw %g3, [%g6 + TI_PRE_COUNT]
#endif
b,pt %xcc, nzero
nop
spin_on_lock:
ldub [%o1], %g3
membar #LoadLoad
brnz,pt %g3, spin_on_lock
nop
ba,pt %xcc, to_zero
nop
nop