netdev - Re: [PATCH net 2/4] tcp: add tcp_rx_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]

Date:   Sun, 16 Jun 2019 15:38:43 +0800
From:   Feng Tang <feng.tang@...el.com>
To:     Eric Dumazet <edumazet@...gle.com>
Cc:     "David S . Miller" <davem@...emloft.net>,
        netdev <netdev@...r.kernel.org>,
        Willem de Bruijn <willemb@...gle.com>,
        Eric Dumazet <eric.dumazet@...il.com>
Subject: Re: [PATCH net 2/4] tcp: add tcp_rx_skb_cache sysctl

On Fri, Jun 14, 2019 at 04:22:19PM -0700, Eric Dumazet wrote:
> Instead of relying on rps_needed, it is safer to use a separate
> static key, since we do not want to enable TCP rx_skb_cache
> by default. This feature can cause huge increase of memory
> usage on hosts with millions of sockets.

Thanks for the effort!

Acked-by: Feng Tang <feng.tang@...el.com>

> 
> Signed-off-by: Eric Dumazet <edumazet@...gle.com>
> ---
>  Documentation/networking/ip-sysctl.txt | 8 ++++++++
>  include/net/sock.h                     | 6 ++----
>  net/ipv4/sysctl_net_ipv4.c             | 9 +++++++++
>  3 files changed, 19 insertions(+), 4 deletions(-)
> 
> diff --git a/Documentation/networking/ip-sysctl.txt b/Documentation/networking/ip-sysctl.txt
> index 14fe93049d28e965d7349b03c5c8782c3d386e7d..288aa264ac26d98637a5bb1babc334bfc699bef1 100644
> --- a/Documentation/networking/ip-sysctl.txt
> +++ b/Documentation/networking/ip-sysctl.txt
> @@ -772,6 +772,14 @@ tcp_challenge_ack_limit - INTEGER
>  	in RFC 5961 (Improving TCP's Robustness to Blind In-Window Attacks)
>  	Default: 100
>  
> +tcp_rx_skb_cache - BOOLEAN
> +	Controls a per TCP socket cache of one skb, that might help
> +	performance of some workloads. This might be dangerous
> +	on systems with a lot of TCP sockets, since it increases
> +	memory usage.
> +
> +	Default: 0 (disabled)
> +
>  UDP variables:
>  
>  udp_l3mdev_accept - BOOLEAN
> diff --git a/include/net/sock.h b/include/net/sock.h
> index e9d769c04637a3c0b967c9bfa6def724834796b9..b02645e2dfad722769c1455bcde76e46da9fc5ac 100644
> --- a/include/net/sock.h
> +++ b/include/net/sock.h
> @@ -2433,13 +2433,11 @@ static inline void skb_setup_tx_timestamp(struct sk_buff *skb, __u16 tsflags)
>   * This routine must be called with interrupts disabled or with the socket
>   * locked so that the sk_buff queue operation is ok.
>  */
> +DECLARE_STATIC_KEY_FALSE(tcp_rx_skb_cache_key);
>  static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
>  {
>  	__skb_unlink(skb, &sk->sk_receive_queue);
> -	if (
> -#ifdef CONFIG_RPS
> -	    !static_branch_unlikely(&rps_needed) &&
> -#endif
> +	if (static_branch_unlikely(&tcp_rx_skb_cache_key) &&
>  	    !sk->sk_rx_skb_cache) {
>  		sk->sk_rx_skb_cache = skb;
>  		skb_orphan(skb);
> diff --git a/net/ipv4/sysctl_net_ipv4.c b/net/ipv4/sysctl_net_ipv4.c
> index 875867b64d6a6597bf4fcd3498ed55741cbe33f7..886b58d31351df44725bdc34081e798bcb89ecf0 100644
> --- a/net/ipv4/sysctl_net_ipv4.c
> +++ b/net/ipv4/sysctl_net_ipv4.c
> @@ -51,6 +51,9 @@ static int comp_sack_nr_max = 255;
>  static u32 u32_max_div_HZ = UINT_MAX / HZ;
>  static int one_day_secs = 24 * 3600;
>  
> +DEFINE_STATIC_KEY_FALSE(tcp_rx_skb_cache_key);
> +EXPORT_SYMBOL(tcp_rx_skb_cache_key);
> +
>  /* obsolete */
>  static int sysctl_tcp_low_latency __read_mostly;
>  
> @@ -559,6 +562,12 @@ static struct ctl_table ipv4_table[] = {
>  		.extra1		= &sysctl_fib_sync_mem_min,
>  		.extra2		= &sysctl_fib_sync_mem_max,
>  	},
> +	{
> +		.procname	= "tcp_rx_skb_cache",
> +		.data		= &tcp_rx_skb_cache_key.key,
> +		.mode		= 0644,
> +		.proc_handler	= proc_do_static_key,
> +	},
>  	{ }
>  };
>  
> -- 
> 2.22.0.410.gd8fdbe21b5-goog