kernel_optimize_test/arch/sparc64/lib/rwsem.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

171 lines
2.9 KiB
ArmAsm

/* rwsem.S: RW semaphore assembler.
*
* Written by David S. Miller (davem@redhat.com), 2001.
* Derived from asm-i386/rwsem.h
*/
#include <asm/rwsem-const.h>
.section .sched.text
.globl __down_read
__down_read:
1: lduw [%o0], %g1
add %g1, 1, %g7
cas [%o0], %g1, %g7
cmp %g1, %g7
bne,pn %icc, 1b
add %g7, 1, %g7
cmp %g7, 0
membar #StoreLoad | #StoreStore
bl,pn %icc, 3f
nop
2:
retl
nop
3:
save %sp, -192, %sp
call rwsem_down_read_failed
mov %i0, %o0
ret
restore
.size __down_read, .-__down_read
.globl __down_read_trylock
__down_read_trylock:
1: lduw [%o0], %g1
add %g1, 1, %g7
cmp %g7, 0
bl,pn %icc, 2f
mov 0, %o1
cas [%o0], %g1, %g7
cmp %g1, %g7
bne,pn %icc, 1b
mov 1, %o1
membar #StoreLoad | #StoreStore
2: retl
mov %o1, %o0
.size __down_read_trylock, .-__down_read_trylock
.globl __down_write
__down_write:
sethi %hi(RWSEM_ACTIVE_WRITE_BIAS), %g1
or %g1, %lo(RWSEM_ACTIVE_WRITE_BIAS), %g1
1:
lduw [%o0], %g3
add %g3, %g1, %g7
cas [%o0], %g3, %g7
cmp %g3, %g7
bne,pn %icc, 1b
cmp %g7, 0
membar #StoreLoad | #StoreStore
bne,pn %icc, 3f
nop
2: retl
nop
3:
save %sp, -192, %sp
call rwsem_down_write_failed
mov %i0, %o0
ret
restore
.size __down_write, .-__down_write
.globl __down_write_trylock
__down_write_trylock:
sethi %hi(RWSEM_ACTIVE_WRITE_BIAS), %g1
or %g1, %lo(RWSEM_ACTIVE_WRITE_BIAS), %g1
1:
lduw [%o0], %g3
cmp %g3, 0
bne,pn %icc, 2f
mov 0, %o1
add %g3, %g1, %g7
cas [%o0], %g3, %g7
cmp %g3, %g7
bne,pn %icc, 1b
mov 1, %o1
membar #StoreLoad | #StoreStore
2: retl
mov %o1, %o0
.size __down_write_trylock, .-__down_write_trylock
.globl __up_read
__up_read:
1:
lduw [%o0], %g1
sub %g1, 1, %g7
cas [%o0], %g1, %g7
cmp %g1, %g7
bne,pn %icc, 1b
cmp %g7, 0
membar #StoreLoad | #StoreStore
bl,pn %icc, 3f
nop
2: retl
nop
3: sethi %hi(RWSEM_ACTIVE_MASK), %g1
sub %g7, 1, %g7
or %g1, %lo(RWSEM_ACTIVE_MASK), %g1
andcc %g7, %g1, %g0
bne,pn %icc, 2b
nop
save %sp, -192, %sp
call rwsem_wake
mov %i0, %o0
ret
restore
.size __up_read, .-__up_read
.globl __up_write
__up_write:
sethi %hi(RWSEM_ACTIVE_WRITE_BIAS), %g1
or %g1, %lo(RWSEM_ACTIVE_WRITE_BIAS), %g1
1:
lduw [%o0], %g3
sub %g3, %g1, %g7
cas [%o0], %g3, %g7
cmp %g3, %g7
bne,pn %icc, 1b
sub %g7, %g1, %g7
cmp %g7, 0
membar #StoreLoad | #StoreStore
bl,pn %icc, 3f
nop
2:
retl
nop
3:
save %sp, -192, %sp
call rwsem_wake
mov %i0, %o0
ret
restore
.size __up_write, .-__up_write
.globl __downgrade_write
__downgrade_write:
sethi %hi(RWSEM_WAITING_BIAS), %g1
or %g1, %lo(RWSEM_WAITING_BIAS), %g1
1:
lduw [%o0], %g3
sub %g3, %g1, %g7
cas [%o0], %g3, %g7
cmp %g3, %g7
bne,pn %icc, 1b
sub %g7, %g1, %g7
cmp %g7, 0
membar #StoreLoad | #StoreStore
bl,pn %icc, 3f
nop
2:
retl
nop
3:
save %sp, -192, %sp
call rwsem_downgrade_wake
mov %i0, %o0
ret
restore
.size __downgrade_write, .-__downgrade_write