Message-ID: <sashiko-outbox-208972@kernel.org>
From: sashiko-bot@kernel.org
Subject: Re: [PATCH bpf-next] bpf, arm32: build constants inline instead of
 using a literal pool
Reply-To: sashiko-reviews@lists.linux.dev
To: "Ramin Moussavi" <ramin.moussavi@yacoub.de>
Cc: bpf@vger.kernel.org
In-Reply-To: <20261111121621.665533-1-ramin.moussavi@yacoub.de>
References: <20261111121621.665533-1-ramin.moussavi@yacoub.de>
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
Date: Wed, 11 Nov 2026 12:23:46 +0000
Precedence: list

Thank you for your contribution! Sashiko AI review found 2 potential
issues to consider:
- [Low] bpf, arm32: non-standard multi-line comment style in
  emit_mov_i_no8m()
- [Low] bpf, arm32: unused flags field left in struct jit_ctx
--

commit 04b7597139a2756fa5ffbad42747333b65860a8c
Author: Ramin Moussavi <ramin.moussavi@yacoub.de>

bpf, arm32: build constants inline instead of using a literal pool

This commit replaces the literal pool used for non-imm8m constants on
ARMv5 and ARMv6 with inline MVN/MOV/ORR instruction sequences and removes
the literal pool tracking fields and overflow fallback.

> diff --git a/arch/arm/net/bpf_jit_32.c b/arch/arm/net/bpf_jit_32.c
> index 9ede81afbc50e..d3b5d77450687 100644
> --- a/arch/arm/net/bpf_jit_32.c
> +++ b/arch/arm/net/bpf_jit_32.c
> @@ -130,8 +130,6 @@ enum {
>  #define TMP_REG_2	(MAX_BPF_JIT_REG + 1)	/* TEMP Register 2 */
>  #define TCALL_CNT	(MAX_BPF_JIT_REG + 2)	/* Tail Call Count */
> =20
> -#define FLAG_IMM_OVERFLOW	(1 << 0)
> -

[ ... ]

> @@ -209,11 +203,6 @@ struct jit_ctx {
>  	u32 *offsets;
>  	u32 *target;
>  	u32 stack_size;
> -#if __LINUX_ARM_ARCH__ < 7
> -	u16 epilogue_bytes;
> -	u16 imm_count;
> -	u32 *imms;
> -#endif
>  };

[Severity: Low]
With FLAG_IMM_OVERFLOW removed, is struct jit_ctx's flags member now
unused as well?  Nothing else in arch/arm/net/bpf_jit_32.c reads or writes
ctx->flags, so it can be dropped alongside epilogue_bytes, imm_count, and
imms.

[ ... ]

> @@ -469,7 +411,30 @@ static inline int bpf2a32_offset(int bpf_to, int bpf=
_from,
>  static inline void emit_mov_i_no8m(const u8 rd, u32 val, struct jit_ctx =
*ctx)
>  {
>  #if __LINUX_ARM_ARCH__ < 7
> -	emit(ARM_LDR_I(rd, ARM_PC, imm_offset(val, ctx)), ctx);
> +	/* A single MVN if the inverted value is an imm8m, else one
> +	 * MOV/ORR per non-zero byte -- a byte is always an imm8m.
> +	 */

[Severity: Low]
This isn't a bug, but BPF code uses the multi-line comment style where
the opening /* sits on its own line and the comment text starts on the
next line.

--=20
Sashiko AI review =C2=B7 https://sashiko.dev/#/patchset/20261111121621.6655=
33-1-ramin.moussavi@yacoub.de?part=3D1