Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 10 May 2018 20:19:16 -0700 (PDT)
From:      "Rodney W. Grimes" <freebsd@pdx.rh.CN85.dnsmgr.net>
To:        =?UTF-8?Q?Dag-Erling_Sm=C3=B8rgrav?= <des@des.no>
Cc:        rgrimes@freebsd.org, src-committers@freebsd.org, svn-src-all@freebsd.org,  svn-src-head@freebsd.org
Subject:   Re: svn commit: r333476 - head/sys/net
Message-ID:  <201805110319.w4B3JGBv073462@pdx.rh.CN85.dnsmgr.net>
In-Reply-To: <86h8nfvtbk.fsf@next.des.no>

next in thread | previous in thread | raw e-mail | index | archive | help
[ Charset UTF-8 unsupported, converting... ]
> "Rodney W. Grimes" <freebsd@pdx.rh.CN85.dnsmgr.net> writes:
> > I do no need or want these routes created by this mechanism on my
> > FreeBSD based routers, they only ever existed originally to use the
> > MTU of the lo0 interface for things that wrongly open an IP address of
> > an interface rather than 127.0.0.1.
> 
> I'm not sure I understand enough of this to make an informed decision.
> Are you saying they serve no practical purpose ever on current systems?
> 
> These are the entries that route all traffic to any of our own addresses
> over lo0, right?  Like the bottom four here?

Correct.
 
> des@hive ~% netstat -4rn | grep -w lo0
> 127.0.0.1          link#2             UH          lo0
> 192.168.144.15     link#1             UHS         lo0
> 192.168.144.16     link#1             UHS         lo0
> 192.168.144.19     link#1             UHS         lo0
> 192.168.144.30     link#1             UHS         lo0
> 
> I'm going to try a kernel with that code #ifdefed out...

We did this back in the day so if you make a connection
to your own IP (your 4 addresses above) then you use
the MTU of lo0, which is usually 16k, instead of the MTU
of the interface, which is usually 1500, or back in the
day of SLIP was often 296.
 
Eugene Grosbein says there is some use case for it where
he has lots of tunX devices, but I just can not seem to
understand why or what it is that he is doing that requires
these routes.

IIRC Linux has never had these routes.  If you go back
in the history of svn and read whey this was originally
added it was to compensate for the fact that you lost the
/etc/rc installed route if you did an if down/up, which
is actually expected behavior, you loose ALL routes
associtated with the IP of an interface if you down it.

The CURRENT code makes it possible for you to ping
the IP of a DOWNED interface due to this now silly
lo0 route, just more demonstration of how wrong
this idea is.  This was caused by the addition of
pinning these routes so that they stay in effect
even when you down the interface.

-- 
Rod Grimes                                                 rgrimes@freebsd.org



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201805110319.w4B3JGBv073462>