netdev - Re: [PATCH v2 5/5] bpf ppc32: Add instructions for atomic

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Sat, 11 Jun 2022 17:34:09 +0000
From:   Christophe Leroy <christophe.leroy@...roup.eu>
To:     Hari Bathini <hbathini@...ux.ibm.com>,
        "bpf@...r.kernel.org" <bpf@...r.kernel.org>,
        linuxppc-dev <linuxppc-dev@...ts.ozlabs.org>
CC:     Michael Ellerman <mpe@...erman.id.au>,
        "Naveen N. Rao" <naveen.n.rao@...ux.ibm.com>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        Benjamin Herrenschmidt <benh@...nel.crashing.org>,
        Paul Mackerras <paulus@...ba.org>,
        Alexei Starovoitov <ast@...nel.org>,
        Daniel Borkmann <daniel@...earbox.net>,
        Andrii Nakryiko <andrii@...nel.org>,
        Martin KaFai Lau <kafai@...com>,
        Song Liu <songliubraving@...com>, Yonghong Song <yhs@...com>,
        John Fastabend <john.fastabend@...il.com>,
        KP Singh <kpsingh@...nel.org>,
        Jordan Niethe <jniethe5@...il.com>,
        Russell Currey <ruscur@...sell.cc>
Subject: Re: [PATCH v2 5/5] bpf ppc32: Add instructions for atomic_[cmp]xchg



Le 10/06/2022 à 17:55, Hari Bathini a écrit :
> This adds two atomic opcodes BPF_XCHG and BPF_CMPXCHG on ppc32, both
> of which include the BPF_FETCH flag.  The kernel's atomic_cmpxchg
> operation fundamentally has 3 operands, but we only have two register
> fields. Therefore the operand we compare against (the kernel's API
> calls it 'old') is hard-coded to be BPF_REG_R0. Also, kernel's
> atomic_cmpxchg returns the previous value at dst_reg + off. JIT the
> same for BPF too with return value put in BPF_REG_0.
> 
>    BPF_REG_R0 = atomic_cmpxchg(dst_reg + off, BPF_REG_R0, src_reg);
> 
> Signed-off-by: Hari Bathini <hbathini@...ux.ibm.com>
> ---
> 
> Changes in v2:
> * Moved variable declaration to avoid late declaration error on
>    some compilers.
> * Tried to make code readable and compact.
> 
> 
>   arch/powerpc/net/bpf_jit_comp32.c | 25 ++++++++++++++++++++++---
>   1 file changed, 22 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/powerpc/net/bpf_jit_comp32.c b/arch/powerpc/net/bpf_jit_comp32.c
> index 28dc6a1a8f2f..43f1c76d48ce 100644
> --- a/arch/powerpc/net/bpf_jit_comp32.c
> +++ b/arch/powerpc/net/bpf_jit_comp32.c
> @@ -297,6 +297,7 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, struct codegen_context *
>   		u32 ax_reg = bpf_to_ppc(BPF_REG_AX);
>   		u32 tmp_reg = bpf_to_ppc(TMP_REG);
>   		u32 size = BPF_SIZE(code);
> +		u32 save_reg, ret_reg;
>   		s16 off = insn[i].off;
>   		s32 imm = insn[i].imm;
>   		bool func_addr_fixed;
> @@ -799,6 +800,9 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, struct codegen_context *
>   		 * BPF_STX ATOMIC (atomic ops)
>   		 */
>   		case BPF_STX | BPF_ATOMIC | BPF_W:
> +			save_reg = _R0;
> +			ret_reg = src_reg;
> +
>   			bpf_set_seen_register(ctx, tmp_reg);
>   			bpf_set_seen_register(ctx, ax_reg);
>   
> @@ -829,6 +833,21 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, struct codegen_context *
>   			case BPF_XOR | BPF_FETCH:
>   				EMIT(PPC_RAW_XOR(_R0, _R0, src_reg));
>   				break;
> +			case BPF_CMPXCHG:
> +				/*
> +				 * Return old value in BPF_REG_0 for BPF_CMPXCHG &
> +				 * in src_reg for other cases.
> +				 */
> +				ret_reg = bpf_to_ppc(BPF_REG_0);
> +
> +				/* Compare with old value in BPF_REG_0 */
> +				EMIT(PPC_RAW_CMPW(bpf_to_ppc(BPF_REG_0), _R0));
> +				/* Don't set if different from old value */
> +				PPC_BCC_SHORT(COND_NE, (ctx->idx + 3) * 4);
> +				fallthrough;
> +			case BPF_XCHG:
> +				save_reg = src_reg;

I'm a bit lost, when save_reg is src_reg, don't we expect the upper part 
(ie src_reg - 1) to be explicitely zeroised ?

> +				break;
>   			default:
>   				pr_err_ratelimited("eBPF filter atomic op code %02x (@%d) unsupported\n",
>   						   code, i);
> @@ -836,15 +855,15 @@ int bpf_jit_build_body(struct bpf_prog *fp, u32 *image, struct codegen_context *
>   			}
>   
>   			/* store new value */
> -			EMIT(PPC_RAW_STWCX(_R0, tmp_reg, dst_reg));
> +			EMIT(PPC_RAW_STWCX(save_reg, tmp_reg, dst_reg));
>   			/* we're done if this succeeded */
>   			PPC_BCC_SHORT(COND_NE, tmp_idx);
>   
>   			/* For the BPF_FETCH variant, get old data into src_reg */
>   			if (imm & BPF_FETCH) {
> -				EMIT(PPC_RAW_MR(src_reg, ax_reg));
> +				EMIT(PPC_RAW_MR(ret_reg, ax_reg));
>   				if (!fp->aux->verifier_zext)
> -					EMIT(PPC_RAW_LI(src_reg_h, 0));
> +					EMIT(PPC_RAW_LI(ret_reg - 1, 0)); /* higher 32-bit */
>   			}
>   			break;
>