On Thu, 15 Dec 2016 11:05:08 +0000 (UTC)
Stuart Henderson <s...@spacehopper.org> wrote:
> On 2016-12-15, Patrick Dohman <patrick_doh...@centurylink.net> wrote:
> > Stuart
> >
> > Please see below for more info:
> >
> > Please note the 5.7 dmesg is subsequent to a reboot.
>
> Thanks. I was wondering about a bug with LCP echoes I accidentally
> introduced that made it into 5.9 (fixed for 6.0).
>
> Nothing stands out from what you've sent. Some possibilities:
>
> - connection somewhere between the APU and the ISP really is dropping
> out (are you using the same cable for the different locations you
> placed the APU in? could a cable be bad? check for errors on the
> ethernet interface)
>
> - machine too busy to handle traffic - maybe tail
> -f /var/log/messages in the background while "vmstat -w 10" or
> something is running (maybe under "script"), look for the timeouts in
> the output and see what cpu is doing at the time
>
> > pass out quick on egress inet6 proto { tcp, udp } from
> > { (pppoe0:network), (athn0:network), (re2:network) } modulate
> > state
>
> btw using (...) causes an extra address lookup to be done when the
> rule is evaluated (i.e. when a packet doesn't match existing state) -
> you may need this for pppoe0 but you can save a bit of cpu with
>
> pass out quick on egress inet6 proto { tcp, udp } from
> { (pppoe0:network), athn0:network, re2:network } modulate state
>
> (and same for the v4 rule)
>
> > ### --- Optional Runtime Options --- ###
> > set optimization conservative
>
> not likely to be the problem, but you're pretty unlikely to need that.
>
Hello Stuart,
I am long time reader of this mailing list (since 5.1 when I started
with OpenBSD). This thread is my first input.
> Thanks. I was wondering about a bug with LCP echoes I accidentally
> introduced that made it into 5.9 (fixed for 6.0).
could you please point me to the changes you are talking about here.
I started using pppoe in 5.9 and the LCP-echo gave me a hard time. I
frequently told my ISP (Deutsche Telekom) to drop the line because I
was hitting the MAXALIVECNT value in if_spppsubr. This happened as soon
as I was in the "lucky" situation that nobody was penetrating me on
ports like ssh, telnet or smtp from outside. I made a modification to
send 'sp->pp_alivecnt' to syslog anytime it was changed in addition to
packet capturing. It turned out that my ISP sends LCP-keepalives in a
45s interval and not every 15s. This means I might eventually get my
first LCP-echo from the provider when pppoe is already timing out. I
would be glad though if there was a way to address this problem without
a custum kernel.
Thanks,
Thomas Braun
