Re: [PATCH bpf-next 2/8] arm32, bpf: add support for sign-extension load instruction

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue, Sep 05, 2023 at 09:06:15PM +0000, Puranjay Mohan wrote:
> The cpuv4 added the support of an instruction that is similar to load
> but also sign-extends the result after the load.
> 
> BPF_MEMSX | <size> | BPF_LDX means dst = *(signed size *) (src + offset)
> here <size> can be one of BPF_B, BPF_H, BPF_W.
> 
> ARM32 has instructions to load a byte or a half word with sign
> extension into a 32bit register. As the JIT uses two 32 bit registers
> to simulate a 64-bit BPF register, an extra instruction is emitted to
> sign-extent the result up to the second register.
> 
> Signed-off-by: Puranjay Mohan <puranjay12@xxxxxxxxx>
> ---
>  arch/arm/net/bpf_jit_32.c | 69 ++++++++++++++++++++++++++++++++++++++-
>  arch/arm/net/bpf_jit_32.h |  2 ++
>  2 files changed, 70 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/arm/net/bpf_jit_32.c b/arch/arm/net/bpf_jit_32.c
> index b26579da770e..f7c162479cf2 100644
> --- a/arch/arm/net/bpf_jit_32.c
> +++ b/arch/arm/net/bpf_jit_32.c
> @@ -333,6 +333,9 @@ static u32 arm_bpf_ldst_imm8(u32 op, u8 rt, u8 rn, s16 imm8)
>  #define ARM_LDRD_I(rt, rn, off)	arm_bpf_ldst_imm8(ARM_INST_LDRD_I, rt, rn, off)
>  #define ARM_LDRH_I(rt, rn, off)	arm_bpf_ldst_imm8(ARM_INST_LDRH_I, rt, rn, off)
>  
> +#define ARM_LDRSH_I(rt, rn, off) arm_bpf_ldst_imm8(ARM_INST_LDRSH_I, rt, rn, off)
> +#define ARM_LDRSB_I(rt, rn, off) arm_bpf_ldst_imm8(ARM_INST_LDRSB_I, rt, rn, off)
> +
>  #define ARM_STR_I(rt, rn, off)	arm_bpf_ldst_imm12(ARM_INST_STR_I, rt, rn, off)
>  #define ARM_STRB_I(rt, rn, off)	arm_bpf_ldst_imm12(ARM_INST_STRB_I, rt, rn, off)
>  #define ARM_STRD_I(rt, rn, off)	arm_bpf_ldst_imm8(ARM_INST_STRD_I, rt, rn, off)
> @@ -1026,6 +1029,24 @@ static bool is_ldst_imm(s16 off, const u8 size)
>  	return -off_max <= off && off <= off_max;
>  }
>  
> +static bool is_ldst_imm8(s16 off, const u8 size)
> +{
> +	s16 off_max = 0;
> +
> +	switch (size) {
> +	case BPF_B:
> +		off_max = 0xff;
> +		break;
> +	case BPF_W:
> +		off_max = 0xfff;
> +		break;
> +	case BPF_H:
> +		off_max = 0xff;
> +		break;
> +	}
> +	return -off_max <= off && off <= off_max;
> +}
> +
>  /* *(size *)(dst + off) = src */
>  static inline void emit_str_r(const s8 dst, const s8 src[],
>  			      s16 off, struct jit_ctx *ctx, const u8 sz){
> @@ -1105,6 +1126,45 @@ static inline void emit_ldx_r(const s8 dst[], const s8 src,
>  	arm_bpf_put_reg64(dst, rd, ctx);
>  }
>  
> +/* dst = *(signed size*)(src + off) */
> +static inline void emit_ldsx_r(const s8 dst[], const s8 src,
> +			       s16 off, struct jit_ctx *ctx, const u8 sz){
> +	const s8 *tmp = bpf2a32[TMP_REG_1];
> +	const s8 *rd = is_stacked(dst_lo) ? tmp : dst;
> +	s8 rm = src;
> +
> +	if (!is_ldst_imm8(off, sz)) {
> +		emit_a32_mov_i(tmp[0], off, ctx);
> +		emit(ARM_ADD_R(tmp[0], tmp[0], src), ctx);

Hmm. This looks inefficient when "off" is able to fit in an immediate.
Please try:

	int add_off;

	if (!is_ldst_imm8(off, sz)) {
		add_off = imm8m(off);
		if (add_off > 0) {
			emit(ARM_ADD_I(tmp[0], src, add_off), ctx);
			rm = tmp[0];
		} else {
			emit_a32_mov_i(tmp[0], off, ctx);
			emit(ARM_ADD_R(tmp[0], tmp[0], src), ctx);
			rm = tmp[0];
		}
		off = 0;
> +	} else if (rd[1] == rm) {
> +		emit(ARM_MOV_R(tmp[0], rm), ctx);
> +		rm = tmp[0];

Why do you need this? rd and rm can be the same for LDRS[BH].

> +	}
> +	switch (sz) {
> +	case BPF_B:
> +		/* Load a Byte with sign extension*/
> +		emit(ARM_LDRSB_I(rd[1], rm, off), ctx);
> +		/* Carry the sign extension to upper 32 bits */
> +		emit(ARM_ASR_I(rd[0], rd[1], 31), ctx);
> +		break;
> +	case BPF_H:
> +		/* Load a HalfWord with sign extension*/
> +		emit(ARM_LDRSH_I(rd[1], rm, off), ctx);
> +		/* Carry the sign extension to upper 32 bits */
> +		emit(ARM_ASR_I(rd[0], rd[1], 31), ctx);
> +		break;
> +	case BPF_W:
> +		/* Load a Word*/
> +		emit(ARM_LDR_I(rd[1], rm, off), ctx);
> +		/* Carry the sign extension to upper 32 bits */
> +		emit(ARM_ASR_I(rd[0], rd[1], 31), ctx);

The last instruction extending to the upper 32 bits is the same in each
of these cases, so is there any reason not to do it outside the switch
statement?

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!




[Index of Archives]     [Linux Samsung SoC]     [Linux Rockchip SoC]     [Linux Actions SoC]     [Linux for Synopsys ARC Processors]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]


  Powered by Linux