From mboxrd@z Thu Jan  1 00:00:00 1970
From: Alexei Starovoitov <ast@fb.com>
Subject: Re: [PATCH bpf 03/11] bpf: Add write access to tcp_sock and sock
 fields
Date: Tue, 19 Dec 2017 17:10:31 -0800
Message-ID: <506567da-0aa5-6e45-a393-7ab3fbad8aa3@fb.com>
References: <20171219062200.372711-1-brakmo@fb.com>
 <20171219062200.372711-4-brakmo@fb.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="windows-1252"; format=flowed
Content-Transfer-Encoding: 7bit
Cc: Kernel Team <kernel-team@fb.com>, Blake Matheny <bmatheny@fb.com>,
        Daniel Borkmann <daniel@iogearbox.net>
To: Lawrence Brakmo <brakmo@fb.com>, netdev <netdev@vger.kernel.org>
Return-path: <netdev-owner@vger.kernel.org>
Received: from mx0b-00082601.pphosted.com ([67.231.153.30]:58118 "EHLO
        mx0b-00082601.pphosted.com" rhost-flags-OK-OK-OK-OK)
        by vger.kernel.org with ESMTP id S1753731AbdLTBKw (ORCPT
        <rfc822;netdev@vger.kernel.org>); Tue, 19 Dec 2017 20:10:52 -0500
In-Reply-To: <20171219062200.372711-4-brakmo@fb.com>
Sender: netdev-owner@vger.kernel.org
List-ID: <netdev.vger.kernel.org>

On 12/18/17 10:21 PM, Lawrence Brakmo wrote:
> +#define SOCK_OPS_SET_FIELD(FIELD_NAME, OBJ)				      \
> +	do {								      \
> +		int reg = BPF_REG_9;					      \
> +		BUILD_BUG_ON(FIELD_SIZEOF(OBJ, FIELD_NAME) >		      \
> +			     FIELD_SIZEOF(struct bpf_sock_ops, FIELD_NAME));  \
> +		while (si->dst_reg == reg || si->src_reg == reg)	      \
> +			reg--;						      \
> +		*insn++ = BPF_STX_MEM(BPF_DW, si->dst_reg, reg,		      \
> +				      offsetof(struct bpf_sock_ops_kern,      \
> +					       temp));			      \
> +		*insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(			      \
> +						struct bpf_sock_ops_kern,     \
> +						is_fullsock),		      \
> +				      reg, si->dst_reg,			      \
> +				      offsetof(struct bpf_sock_ops_kern,      \
> +					       is_fullsock));		      \
> +		*insn++ = BPF_JMP_IMM(BPF_JEQ, reg, 0, 2);		      \
> +		*insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(			      \
> +						struct bpf_sock_ops_kern, sk),\
> +				      reg, si->dst_reg,			      \
> +				      offsetof(struct bpf_sock_ops_kern, sk));\
> +		*insn++ = BPF_STX_MEM(BPF_FIELD_SIZEOF(OBJ, FIELD_NAME),      \
> +				      reg, si->src_reg,			      \
> +				      offsetof(OBJ, FIELD_NAME));	      \
> +		*insn++ = BPF_LDX_MEM(BPF_DW, reg, si->dst_reg,		      \
> +				      offsetof(struct bpf_sock_ops_kern,      \
> +					       temp));			      \
> +	} while (0)

that's neat. I like it.
I guess the prog can check is_fullsock on its own to see whether writes
will fail or not, so JEQ above is ok.
Only while() loop looks a bit scary.
May be replace with two 'if' ?
if (si->dst_reg == reg || si->src_reg == reg)
   reg --;
if (si->dst_reg == reg || si->src_reg == reg)
   reg --;
so it's clear that tmp reg will be reg_7, 8 or 9.