After converting arm64 to Generic Entry framework, the compiler no longer
inlines el0_svc_common() into its caller do_el0_svc(). This introduces
a small but measurable overhead in the critical system call path.
Manually forcing el0_svc_common() to be inlined restores the
performance. Benchmarking with perf bench syscall basic on a
Kunpeng 920 platform (based on v6.19-rc1) shows a ~1% performance
uplift.
Inlining this function reduces function prologue/epilogue overhead
and allows for better compiler optimization in the hot system call
dispatch path.
| Metric | W/O this patch | With this patch | Change |
| ---------- | -------------- | --------------- | --------- |
| Total time | 2.195 [sec] | 2.171 [sec] | ↓1.1% |
| usecs/op | 0.219575 | 0.217192 | ↓1.1% |
| ops/sec | 4,554,260 | 4,604,225 | ↑1.1% |
Reviewed-by: Kevin Brodsky <kevin.brodsky@arm.com>
Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
---
arch/arm64/kernel/syscall.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/arch/arm64/kernel/syscall.c b/arch/arm64/kernel/syscall.c
index 77d00a5cf0e9..6fcd97c46716 100644
--- a/arch/arm64/kernel/syscall.c
+++ b/arch/arm64/kernel/syscall.c
@@ -66,8 +66,8 @@ static void invoke_syscall(struct pt_regs *regs, unsigned int scno,
choose_random_kstack_offset(get_random_u16());
}
-static void el0_svc_common(struct pt_regs *regs, int scno, int sc_nr,
- const syscall_fn_t syscall_table[])
+static __always_inline void el0_svc_common(struct pt_regs *regs, int scno, int sc_nr,
+ const syscall_fn_t syscall_table[])
{
unsigned long work = READ_ONCE(current_thread_info()->syscall_work);
unsigned long flags = read_thread_flags();
--
2.34.1