linux-stable/arch/x86/lib/atomic64_cx8_32.S
Uros Bizjak 6e30a7c98a locking/atomic/x86: Introduce the read64_nonatomic macro to x86_32 with cx8
As described in commit:

  e73c4e34a0 ("locking/atomic/x86: Introduce arch_atomic64_read_nonatomic() to x86_32")

the value preload before the CMPXCHG loop does not need to be atomic.

Introduce the read64_nonatomic assembly macro to load the value from a
atomic_t location in a faster non-atomic way and use it in
atomic64_cx8_32.S.

Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: https://lore.kernel.org/r/20240605181424.3228-1-ubizjak@gmail.com
2024-07-17 16:28:11 +02:00

186 lines
2.7 KiB
ArmAsm

/* SPDX-License-Identifier: GPL-2.0-or-later */
/*
* atomic64_t for 586+
*
* Copyright © 2010 Luca Barbieri
*/
#include <linux/linkage.h>
#include <asm/alternative.h>
.macro read64 reg
movl %ebx, %eax
movl %ecx, %edx
/* we need LOCK_PREFIX since otherwise cmpxchg8b always does the write */
LOCK_PREFIX
cmpxchg8b (\reg)
.endm
.macro read64_nonatomic reg
movl (\reg), %eax
movl 4(\reg), %edx
.endm
SYM_FUNC_START(atomic64_read_cx8)
read64 %ecx
RET
SYM_FUNC_END(atomic64_read_cx8)
SYM_FUNC_START(atomic64_set_cx8)
1:
/* we don't need LOCK_PREFIX since aligned 64-bit writes
* are atomic on 586 and newer */
cmpxchg8b (%esi)
jne 1b
RET
SYM_FUNC_END(atomic64_set_cx8)
SYM_FUNC_START(atomic64_xchg_cx8)
1:
LOCK_PREFIX
cmpxchg8b (%esi)
jne 1b
RET
SYM_FUNC_END(atomic64_xchg_cx8)
.macro addsub_return func ins insc
SYM_FUNC_START(atomic64_\func\()_return_cx8)
pushl %ebp
pushl %ebx
pushl %esi
pushl %edi
movl %eax, %esi
movl %edx, %edi
movl %ecx, %ebp
read64_nonatomic %ecx
1:
movl %eax, %ebx
movl %edx, %ecx
\ins\()l %esi, %ebx
\insc\()l %edi, %ecx
LOCK_PREFIX
cmpxchg8b (%ebp)
jne 1b
10:
movl %ebx, %eax
movl %ecx, %edx
popl %edi
popl %esi
popl %ebx
popl %ebp
RET
SYM_FUNC_END(atomic64_\func\()_return_cx8)
.endm
addsub_return add add adc
addsub_return sub sub sbb
.macro incdec_return func ins insc
SYM_FUNC_START(atomic64_\func\()_return_cx8)
pushl %ebx
read64_nonatomic %esi
1:
movl %eax, %ebx
movl %edx, %ecx
\ins\()l $1, %ebx
\insc\()l $0, %ecx
LOCK_PREFIX
cmpxchg8b (%esi)
jne 1b
10:
movl %ebx, %eax
movl %ecx, %edx
popl %ebx
RET
SYM_FUNC_END(atomic64_\func\()_return_cx8)
.endm
incdec_return inc add adc
incdec_return dec sub sbb
SYM_FUNC_START(atomic64_dec_if_positive_cx8)
pushl %ebx
read64 %esi
1:
movl %eax, %ebx
movl %edx, %ecx
subl $1, %ebx
sbb $0, %ecx
js 2f
LOCK_PREFIX
cmpxchg8b (%esi)
jne 1b
2:
movl %ebx, %eax
movl %ecx, %edx
popl %ebx
RET
SYM_FUNC_END(atomic64_dec_if_positive_cx8)
SYM_FUNC_START(atomic64_add_unless_cx8)
pushl %ebp
pushl %ebx
/* these just push these two parameters on the stack */
pushl %edi
pushl %ecx
movl %eax, %ebp
movl %edx, %edi
read64 %esi
1:
cmpl %eax, 0(%esp)
je 4f
2:
movl %eax, %ebx
movl %edx, %ecx
addl %ebp, %ebx
adcl %edi, %ecx
LOCK_PREFIX
cmpxchg8b (%esi)
jne 1b
movl $1, %eax
3:
addl $8, %esp
popl %ebx
popl %ebp
RET
4:
cmpl %edx, 4(%esp)
jne 2b
xorl %eax, %eax
jmp 3b
SYM_FUNC_END(atomic64_add_unless_cx8)
SYM_FUNC_START(atomic64_inc_not_zero_cx8)
pushl %ebx
read64 %esi
1:
movl %eax, %ecx
orl %edx, %ecx
jz 3f
movl %eax, %ebx
xorl %ecx, %ecx
addl $1, %ebx
adcl %edx, %ecx
LOCK_PREFIX
cmpxchg8b (%esi)
jne 1b
movl $1, %eax
3:
popl %ebx
RET
SYM_FUNC_END(atomic64_inc_not_zero_cx8)