netdev - Re: [Patch bpf-next 03/19] skmsg: use skb ext instead of TCP_SKB

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: Windows password security audit tool. GUI, reports in PDF.

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

Message-ID: <87eehu4157.fsf@cloudflare.com>
Date:   Fri, 05 Feb 2021 23:09:56 +0100
From:   Jakub Sitnicki <jakub@...udflare.com>
To:     Cong Wang <xiyou.wangcong@...il.com>
Cc:     netdev@...r.kernel.org, bpf@...r.kernel.org,
        duanxiongchun@...edance.com, wangdongdong.6@...edance.com,
        jiang.wang@...edance.com, Cong Wang <cong.wang@...edance.com>,
        John Fastabend <john.fastabend@...il.com>,
        Daniel Borkmann <daniel@...earbox.net>,
        Lorenz Bauer <lmb@...udflare.com>
Subject: Re: [Patch bpf-next 03/19] skmsg: use skb ext instead of TCP_SKB_CB

On Wed, Feb 03, 2021 at 05:16 AM CET, Cong Wang wrote:
> From: Cong Wang <cong.wang@...edance.com>
>
> Currently TCP_SKB_CB() is hard-coded in skmsg code, it certainly
> won't work for any other non-TCP protocols. We can move them to
> skb ext instead of playing with skb cb, which is harder to make
> correct.
>
> Of course, except ->data_end, which is used by
> sk_skb_convert_ctx_access() to adjust compile-time constant offset.
> Fortunately, we can reuse the anonymous union where the field
> 'tcp_tsorted_anchor' is and save/restore the overwritten part
> before/after a brief use.
>
> Cc: John Fastabend <john.fastabend@...il.com>
> Cc: Daniel Borkmann <daniel@...earbox.net>
> Cc: Jakub Sitnicki <jakub@...udflare.com>
> Cc: Lorenz Bauer <lmb@...udflare.com>
> Signed-off-by: Cong Wang <cong.wang@...edance.com>
> ---
>  include/linux/skbuff.h |  4 ++++
>  include/linux/skmsg.h  | 45 ++++++++++++++++++++++++++++++++++++++++++
>  include/net/tcp.h      | 25 -----------------------
>  net/Kconfig            |  1 +
>  net/core/filter.c      |  3 +--
>  net/core/skbuff.c      |  7 +++++++
>  net/core/skmsg.c       | 44 ++++++++++++++++++++++++++++-------------
>  net/core/sock_map.c    | 12 +++++------
>  8 files changed, 94 insertions(+), 47 deletions(-)
>
> diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
> index 46f901adf1a8..12a28268233a 100644
> --- a/include/linux/skbuff.h
> +++ b/include/linux/skbuff.h
> @@ -755,6 +755,7 @@ struct sk_buff {
>  			void		(*destructor)(struct sk_buff *skb);
>  		};
>  		struct list_head	tcp_tsorted_anchor;
> +		void			*data_end;
>  	};

I think we can avoid `data_end` by computing it in BPF with the help of
a scratch register. Similar to how we compute skb_shinfo(skb) in
bpf_convert_shinfo_access(). Something like:

static struct bpf_insn *bpf_convert_data_end_access(const struct bpf_insn *si,
						    struct bpf_insn *insn)
{
	/* dst_reg = skb->data */
	*insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(struct sk_buff, data),
			      si->dst_reg, si->src_reg,
			      offsetof(struct sk_buff, data));
	/* AX = skb->len */
	*insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(struct sk_buff, len),
			      BPF_REG_AX, si->src_reg,
			      offsetof(struct sk_buff, len));
	/* dst_reg = skb->data + skb->len */
	*insn++ = BPF_ALU64_REG(BPF_ADD, si->dst_reg, BPF_REG_AX);
	/* AX = skb->data_len */
	*insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(struct sk_buff, data_len),
			      BPF_REG_AX, si->src_reg,
			      offsetof(struct sk_buff, data_len));
	/* dst_reg = skb->data + skb->len - skb->data_len */
	*insn++ = BPF_ALU64_REG(BPF_SUB, si->dst_reg, BPF_REG_AX);

	return insn;
}

[...]