From mboxrd@z Thu Jan  1 00:00:00 1970
From: Daniel Borkmann <dborkman@redhat.com>
Subject: Re: [PATCH net-next] net: filter: constify detection of pkt_type_offset
Date: Fri, 12 Sep 2014 12:55:15 +0200
Message-ID: <5412D113.6000700@redhat.com>
References: <09acea1aba100a62551605ba52ae8541e816d1e6.1410517225.git.hannes@stressinduktion.org>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Cc: netdev@vger.kernel.org, Eric Dumazet <eric.dumazet@gmail.com>,
	Markos Chandras <markos.chandras@imgtec.com>,
	Martin Schwidefsky <schwidefsky@de.ibm.com>,
	Alexei Starovoitov <alexei.starovoitov@gmail.com>,
	Denis Kirjanov <kda@linux-powerpc.org>
To: Hannes Frederic Sowa <hannes@stressinduktion.org>
Return-path: <netdev-owner@vger.kernel.org>
Received: from mx1.redhat.com ([209.132.183.28]:35761 "EHLO mx1.redhat.com"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1753500AbaILKzd (ORCPT <rfc822;netdev@vger.kernel.org>);
	Fri, 12 Sep 2014 06:55:33 -0400
In-Reply-To: <09acea1aba100a62551605ba52ae8541e816d1e6.1410517225.git.hannes@stressinduktion.org>
Sender: netdev-owner@vger.kernel.org
List-ID: <netdev.vger.kernel.org>

On 09/12/2014 12:22 PM, Hannes Frederic Sowa wrote:
> Currently we have 2 pkt_type_offset functions doing the same thing and
> spread across the architecture files. Remove those and replace them
> with a PKT_TYPE_OFFSET macro helper which gets the constant value from a
> zero sized sk_buff member right in front of the bitfield with offsetof.
> This new offset marker does not change size of struct sk_buff.
>
> Cc: Eric Dumazet <eric.dumazet@gmail.com>
> Cc: Markos Chandras <markos.chandras@imgtec.com>
> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
> Cc: Daniel Borkmann <dborkman@redhat.com>
> Cc: Alexei Starovoitov <alexei.starovoitov@gmail.com>
> Signed-off-by: Denis Kirjanov <kda@linux-powerpc.org>
> Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>

Thanks for doing this, Hannes!
...
>   static u64 __skb_get_pay_offset(u64 ctx, u64 a, u64 x, u64 r4, u64 r5)
>   {
>   	return skb_get_poff((struct sk_buff *)(unsigned long) ctx);
> @@ -191,7 +167,7 @@ static bool convert_bpf_extensions(struct sock_filter *fp,
>
>   	case SKF_AD_OFF + SKF_AD_PKTTYPE:
>   		*insn = BPF_LDX_MEM(BPF_B, BPF_REG_A, BPF_REG_CTX,
> -				    pkt_type_offset());
> +				    PKT_TYPE_OFFSET());
>   		if (insn->off < 0)
>   			return false;
>   		insn++;

I think this can now be rewritten as ...

	case SKF_AD_OFF + SKF_AD_PKTTYPE:
		*insn++ = BPF_LDX_MEM(BPF_B, BPF_REG_A, BPF_REG_CTX,
				      PKT_TYPE_OFFSET());
		*insn = BPF_ALU32_IMM(BPF_AND, BPF_REG_A, PKT_TYPE_MAX);

... since we are guaranteed to have always >0 insn->off and despite the
__s16 type in off we might be able to safely assume that sk_buff will
never get extended so far w/o fist fights first, where this could wrap
around. ;) Alternatively, perhaps a BUILD_BUG_ON() for really being
paranoid?