From: Jinjie Ruan <ruanjinjie@huawei.com>
Date: Mon, 1 Dec 2025 12:06:33 +0000 (+0800)
Subject: arm64: Avoid memcpy() for syscall_get_arguments()
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=98cc091262effcdbdeb1c2ee0a25981aefd2274c;p=thirdparty%2Fkernel%2Flinux.git

arm64: Avoid memcpy() for syscall_get_arguments()

Do not use memcpy() to extract syscall arguments from struct pt_regs
but rather just perform direct assignments.

Update syscall_set_arguments() too to keep syscall_get_arguments()
and syscall_set_arguments() in sync.

With Generic Entry patch[1] and turn on audit, the performance
benchmarks from perf bench basic syscall on kunpeng920 gives roughly
a 1% performance uplift.

| Metric     | W/O this patch | With this patch | Change    |
| ---------- | -------------- | --------------- | --------- |
| Total time | 2.241 [sec]    | 2.211 [sec]     |  ↓1.36%   |
| usecs/op   | 0.224157       | 0.221146        |  ↓1.36%   |
| ops/sec    | 4,461,157      | 4,501,409       |  ↑0.9%    |

Disassembly shows that using direct assignment causes
syscall_set_arguments() to be inlined and cuts the instruction count by
five or six compared to memcpy(). Because __audit_syscall_entry() only
uses four syscall arguments, the compiler has also elided the copy of
regs->regs[4] and regs->regs[5].

Before:
<syscall_get_arguments.constprop.0>:
       aa0103e2        mov     x2, x1
       91002003        add     x3, x0, #0x8
       f9408804        ldr     x4, [x0, #272]
       f8008444        str     x4, [x2], #8
       a9409404        ldp     x4, x5, [x0, #8]
       a9009424        stp     x4, x5, [x1, #8]
       a9418400        ldp     x0, x1, [x0, #24]
       a9010440        stp     x0, x1, [x2, #16]
       f9401060        ldr     x0, [x3, #32]
       f9001040        str     x0, [x2, #32]
       d65f03c0        ret
       d503201f        nop

After:
       a9408e82        ldp     x2, x3, [x20, #8]
       2a1603e0        mov     w0, w22
       f9400e84        ldr     x4, [x20, #24]
       f9408a81        ldr     x1, [x20, #272]
       9401c4ba        bl      ffff800080215ca8 <__audit_syscall_entry>

This also aligns the implementation with x86 and RISC-V.

[1]: https://lore.kernel.org/all/20251126071446.3234218-1-ruanjinjie@huawei.com/

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
Signed-off-by: Will Deacon <will@kernel.org>
---

diff --git a/arch/arm64/include/asm/syscall.h b/arch/arm64/include/asm/syscall.h
index f3853047c28e0..5e4c7fc44f731 100644
--- a/arch/arm64/include/asm/syscall.h
+++ b/arch/arm64/include/asm/syscall.h
@@ -82,16 +82,24 @@ static inline void syscall_get_arguments(struct task_struct *task,
 					 unsigned long *args)
 {
 	args[0] = regs->orig_x0;
-	args++;
-
-	memcpy(args, &regs->regs[1], 5 * sizeof(args[0]));
+	args[1] = regs->regs[1];
+	args[2] = regs->regs[2];
+	args[3] = regs->regs[3];
+	args[4] = regs->regs[4];
+	args[5] = regs->regs[5];
 }
 
 static inline void syscall_set_arguments(struct task_struct *task,
 					 struct pt_regs *regs,
 					 const unsigned long *args)
 {
-	memcpy(&regs->regs[0], args, 6 * sizeof(args[0]));
+	regs->regs[0] = args[0];
+	regs->regs[1] = args[1];
+	regs->regs[2] = args[2];
+	regs->regs[3] = args[3];
+	regs->regs[4] = args[4];
+	regs->regs[5] = args[5];
+
 	/*
 	 * Also copy the first argument into orig_x0
 	 * so that syscall_get_arguments() would return it