svn commit: r215791 - stable/8/sys/netinet

[Bruce Evans]

On Wed, 24 Nov 2010, Gleb Smirnoff wrote:

> Log:
>  MFhead r214508:
>    Revert a small part of the r198301, that is entirely unrelated to the
>    r198301 itself. It also broke the logic of not sending more than one
>    ARP request per second, that consequently lead to a potential problem
>    of flooding network with broadcast packets.
>
> Modified:
>  stable/8/sys/netinet/if_ether.c
>
> Modified: stable/8/sys/netinet/if_ether.c
> ==============================================================================
> --- stable/8/sys/netinet/if_ether.c	Wed Nov 24 05:24:36 2010	(r215790)
> +++ stable/8/sys/netinet/if_ether.c	Wed Nov 24 05:37:12 2010	(r215791)
> @@ -381,7 +381,7 @@ retry:
> 		int canceled;
>
> 		LLE_ADDREF(la);
> -		la->la_expire = time_second + V_arpt_down;
> +		la->la_expire = time_second;
> 		canceled = callout_reset(&la->la_timer, hz * V_arpt_down,
> 		    arptimer, la);
> 		if (canceled)
>

Isn't using non-monotic time for timeouts always wrong?

There are lots of other time_second's in networkining code.  These
still outnumber time_uptime's by about 68:41.  rtcock.c uses the weird
expression time_second - time_uptime for metrics.  Since time_uptime
is relative to boot time while time_second is relative to the Epoch,
their difference is approximately the number of seconds since the
Epoch, which is a very strange value which might nevertheless be
useful for converting between monotonic expiry times and real expiry
times, but I think it doesn't work even for that if the real time is
stepped.

Bruce