netdev - Re: [net-next 1/3] tcp: avoid useless copying and collapsing of just one 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: <1507995761.31614.56.camel@edumazet-glaptop3.roam.corp.google.com>
Date:   Sat, 14 Oct 2017 08:42:41 -0700
From:   Eric Dumazet <eric.dumazet@...il.com>
To:     Koichiro Den <den@...ipeden.com>
Cc:     netdev@...r.kernel.org, davem@...emloft.net, kuznet@....inr.ac.ru,
        yoshfuji@...ux-ipv6.org
Subject: Re: [net-next 1/3] tcp: avoid useless copying and collapsing of
 just one skb

On Sat, 2017-10-14 at 16:27 +0900, Koichiro Den wrote:
> On the starting point chosen, it could be possible that just one skb
> remains in between the range provided, leading to copying and re-insertion
> of rb node, which is useless with respect to the rcv buf measurement.
> This is rather probable in ooo queue case, in which non-contiguous bloated
> packets have been queued up.
> 
> Signed-off-by: Koichiro Den <den@...ipeden.com>
> ---
>  net/ipv4/tcp_input.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
> index d0682ce2a5d6..1d785b5bf62d 100644
> --- a/net/ipv4/tcp_input.c
> +++ b/net/ipv4/tcp_input.c
> @@ -4807,7 +4807,8 @@ tcp_collapse(struct sock *sk, struct sk_buff_head *list, struct rb_root *root,
>  		start = TCP_SKB_CB(skb)->end_seq;
>  	}
>  	if (end_of_skbs ||
> -	    (TCP_SKB_CB(skb)->tcp_flags & (TCPHDR_SYN | TCPHDR_FIN)))
> +	    (TCP_SKB_CB(skb)->tcp_flags & (TCPHDR_SYN | TCPHDR_FIN)) ||
> +	    (TCP_SKB_CB(skb)->seq == start && TCP_SKB_CB(skb)->end_seq == end))
>  		return;
>  
>  	__skb_queue_head_init(&tmp);


What do you mean by useless ?

Surely if this skb contains 17 segments (some USB drivers allocate 8KB
per frame), we want to collapse them to save memory.

So I do not agree with this patch.