Re: [net-next PATCH v3 2/3] net: TCP thin linear timeouts

[William Allen Simpson]

Last year, I'm pretty sure I was on record as thinking this is only a
marginally good idea, that would be better at the application layer.

Also that naming was a bit dicey.  Now the names are more descriptive,
but the "force" is a bit overkill.

How about?
  NET_TCP_FORCE_THIN_LINEAR_TIMEOUTS    -> NET_TCP_THIN_LINEAR_TIMEOUTS
  TCP_THIN_LT                           -> TCP_THIN_LINEAR_TIMEOUTS
  TCP_THIN_LT_RETRIES                   -> TCP_THIN_LINEAR_RETRIES
  tcp_force_thin_linear_timeouts        -> tcp_thin_linear_timeouts
  sysctl_tcp_force_thin_linear_timeouts -> sysctl_tcp_thin_linear_timeouts
  tp->thin_lt                           -> tp->thin_lto

The latter mostly traditional "to" for "timeout", as used most everywhere.


Andreas Petlund wrote:
> diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c
> index de7d1bf..a682479 100644
> --- a/net/ipv4/tcp_timer.c
> +++ b/net/ipv4/tcp_timer.c
> @@ -29,6 +29,7 @@ int sysctl_tcp_keepalive_intvl __read_mostly = TCP_KEEPALIVE_INTVL;
>  int sysctl_tcp_retries1 __read_mostly = TCP_RETR1;
>  int sysctl_tcp_retries2 __read_mostly = TCP_RETR2;
>  int sysctl_tcp_orphan_retries __read_mostly;
> +int sysctl_tcp_force_thin_linear_timeouts __read_mostly;
>  
Where is the sysctl initialized?


> +	if ((tp->thin_lt || sysctl_tcp_force_thin_linear_timeouts) &&
> +	    tcp_stream_is_thin(sk) && sk->sk_state == TCP_ESTABLISHED &&
> +	    icsk->icsk_retransmits <= TCP_THIN_LT_RETRIES) {

Just for efficiency, I'd reorder this
  +	if (sk->sk_state == TCP_ESTABLISHED &&
  +	    (tp->thin_lt || sysctl_tcp_force_thin_linear_timeouts) &&
  +	    tcp_stream_is_thin(sk) &&
  +	    icsk->icsk_retransmits <= TCP_THIN_LT_RETRIES) {

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/