Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <1655275079-17651-1-git-send-email-yangtiezhu@loongson.cn> <1655275079-17651-3-git-send-email-yangtiezhu@loongson.cn>
In-Reply-To: <1655275079-17651-3-git-send-email-yangtiezhu@loongson.cn>
From:   hev <r@hev.cc>
Date:   Wed, 15 Jun 2022 14:57:17 +0800
Message-ID: <CAHirt9g+0N6hJAf-xRrss-FRXiix4jE9=CGZC6+BDHO3dBunsg@mail.gmail.com>
Subject: Re: [RFC PATCH v2 2/2] LoongArch: No need to call RESTORE_ALL_AND_RET
 for all syscalls
To:     Tiezhu Yang <yangtiezhu@loongson.cn>
Cc:     Huacai Chen <chenhuacai@kernel.org>,
        WANG Xuerui <kernel@xen0n.name>,
        Xuefeng Li <lixuefeng@loongson.cn>,
        Jianmin Lv <lvjianmin@loongson.cn>, Jun Yi <yijun@loongson.cn>,
        Rui Wang <wangrui@loongson.cn>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

Hello,

On Wed, Jun 15, 2022 at 2:38 PM Tiezhu Yang <yangtiezhu@loongson.cn> wrote:
>
> In handle_syscall, it is unnecessary to call RESTORE_ALL_AND_RET
> for all syscalls.
>
> (1) If syscall number is __NR_clone and __NR_clone3,
>     call RESTORE_STATIC_SOME_SP_AND_RET.
> (2) If syscall number is __NR_rt_sigreturn and __NR_rt_sigsuspend,
>     call RESTORE_TEMP_SOME_SP_AND_RET.
> (3) The other syscalls call RESTORE_SOME_SP_AND_RET.
>
> With this patch, it can reduce many load instructions.
>
> Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
> ---
>  arch/loongarch/include/asm/stackframe.h | 17 +++++++++++++++++
>  arch/loongarch/kernel/entry.S           | 24 +++++++++++++++++++++++-
>  2 files changed, 40 insertions(+), 1 deletion(-)
>
> diff --git a/arch/loongarch/include/asm/stackframe.h b/arch/loongarch/include/asm/stackframe.h
> index 4ca9530..52649a5f 100644
> --- a/arch/loongarch/include/asm/stackframe.h
> +++ b/arch/loongarch/include/asm/stackframe.h
> @@ -216,4 +216,21 @@
>         RESTORE_SP_AND_RET \docfi
>         .endm
>
> +       .macro  RESTORE_SOME_SP_AND_RET docfi=0
> +       RESTORE_SOME \docfi
> +       RESTORE_SP_AND_RET \docfi
> +       .endm
> +
> +       .macro  RESTORE_STATIC_SOME_SP_AND_RET docfi=0
> +       RESTORE_STATIC \docfi
> +       RESTORE_SOME \docfi
> +       RESTORE_SP_AND_RET \docfi
> +       .endm
> +
> +       .macro  RESTORE_TEMP_SOME_SP_AND_RET docfi=0
> +       RESTORE_TEMP \docfi
> +       RESTORE_SOME \docfi
> +       RESTORE_SP_AND_RET \docfi
> +       .endm
> +
>  #endif /* _ASM_STACKFRAME_H */
> diff --git a/arch/loongarch/kernel/entry.S b/arch/loongarch/kernel/entry.S
> index 53ce2cb..58fe507 100644
> --- a/arch/loongarch/kernel/entry.S
> +++ b/arch/loongarch/kernel/entry.S
> @@ -79,7 +79,29 @@ SYM_FUNC_START(handle_syscall)
>         move    a0, sp
>         bl      do_syscall
>
> -       RESTORE_ALL_AND_RET
> +       /*
> +        * Syscall number held in a7 which is stored in PT_R11.
> +        * If syscall number is __NR_clone and __NR_clone3,
> +        * call RESTORE_STATIC_SOME_SP_AND_RET.
> +        * If syscall number is __NR_rt_sigreturn and __NR_rt_sigsuspend,
> +        * call RESTORE_TEMP_SOME_SP_AND_RET.
> +        * The other syscalls call RESTORE_SOME_SP_AND_RET.
> +        */
> +       cfi_ld  t3, PT_R11

I think PT_R11 may be overwritten by the signal handler and the
syscall number is now lost.

> +       li.w    t4, __NR_clone
> +       beq     t3, t4, 3f
> +       li.w    t4, __NR_clone3
> +       beq     t3, t4, 3f
> +       li.w    t4, __NR_rt_sigreturn
> +       beq     t3, t4, 4f
> +       li.w    t4, __NR_rt_sigsuspend
> +       beq     t3, t4, 4f
> +
> +       RESTORE_SOME_SP_AND_RET
> +3:
> +       RESTORE_STATIC_SOME_SP_AND_RET
> +4:
> +       RESTORE_TEMP_SOME_SP_AND_RET
>  SYM_FUNC_END(handle_syscall)
>
>  SYM_CODE_START(ret_from_fork)
> --
> 2.1.0
>

Best regards,
hev