On 05/31/2017 08:19 AM, James Hogan wrote:
Adjust the atomic loop in the MIPS_ATOMIC_SET operation of the sysmips
system call to branch straight back to the linked load rather than
jumping via a different subsection (whose purpose remains a mystery to
me).
The subsection keeps the code for the (hopefully) cold path out of line
which should result in a smaller cache footprint in the hot path.
Signed-off-by: James Hogan <james.hogan@xxxxxxxxxx>
Cc: Ralf Baechle <ralf@xxxxxxxxxxxxxx>
Cc: linux-mips@xxxxxxxxxxxxxx
---
arch/mips/kernel/syscall.c | 6 +-----
1 file changed, 1 insertion(+), 5 deletions(-)
diff --git a/arch/mips/kernel/syscall.c b/arch/mips/kernel/syscall.c
index ca54ac40252b..6c6bf43d681b 100644
--- a/arch/mips/kernel/syscall.c
+++ b/arch/mips/kernel/syscall.c
@@ -137,13 +137,9 @@ static inline int mips_atomic_set(unsigned long addr, unsigned long new)
" move %[tmp], %[new] \n"
"2: \n"
user_sc("%[tmp]", "(%[addr])")
- " beqz %[tmp], 4f \n"
+ " beqz %[tmp], 1b \n"
"3: \n"
" .insn \n"
- " .subsection 2 \n"
- "4: b 1b \n"
- " .previous \n"
- " \n"
" .section .fixup,\"ax\" \n"
"5: li %[err], %[efault] \n"
" j 3b \n"
