Re: [PATCH] tcp: Fix sysctl_tcp_max_orphans when PAGE_SIZE != 4k

[Eric Dumazet <eric.dumazet@gmail.com>]

Le mercredi 25 août 2010 à 17:17 +1000, Anton Blanchard a écrit :
> We were hard coding 4096 when sizing sysctl_tcp_max_orphans which
> causes problems when PAGE_SIZE is not 4k. We calculate an order based on
> PAGE_SHIFT so the count should be based on PAGE_SIZE
> 
> Signed-off-By: Milton Miller <miltonm@bga.com>
> Signed-off-by: Anton Blanchard <anton@samba.org>
> ---
> 
> Index: powerpc.git/net/ipv4/tcp.c
> ===================================================================
> --- powerpc.git.orig/net/ipv4/tcp.c	2010-08-25 17:04:51.190305401 +1000
> +++ powerpc.git/net/ipv4/tcp.c	2010-08-25 17:05:15.463884764 +1000
> @@ -3270,7 +3270,7 @@ void __init tcp_init(void)
>  		;
>  	if (order >= 4) {
>  		tcp_death_row.sysctl_max_tw_buckets = 180000;
> -		sysctl_tcp_max_orphans = 4096 << (order - 4);
> +		sysctl_tcp_max_orphans = PAGE_SIZE << (order - 4);
>  		sysctl_max_syn_backlog = 1024;
>  	} else if (order < 3) {
>  		tcp_death_row.sysctl_max_tw_buckets >>= (3 - order);
> --

In fact, existing code makes litle sense....

(tcp_hashinfo.bhash_size * sizeof(struct inet_bind_hashbucket))

is much bigger if spinlock debugging is on. Its strange to select bigger
limits in this case (where kernel structures are also bigger)

bhash_size max is 65536, and we get this value even for small machines. 

Sizing would probably be better if using ehash_size instead of
bhash_size

Maybe remove the 'order' loop and use ehash_size, already a result of
the available memory or thash_entries tunable.

unsigned int ehash_size = tcp_hashinfo.ehash_mask + 1;

tcp_death_row.sysctl_max_tw_buckets = cnt / 2;
sysctl_tcp_max_orphans = cnt / 2;
sysctl_max_syn_backlog = min(128, cnt / 256);