Netdev List
 help / color / mirror / Atom feed
From: David Madore <david+ml@madore.org>
To: Linux Netdev Mailing-List <netdev@vger.kernel.org>
Subject: possible routing table cache bug for mtu/advmss attributes in Linux 3.0.4
Date: Thu, 1 Sep 2011 23:46:43 +0200	[thread overview]
Message-ID: <20110901214643.GA16989@aldebaran.gro-tsen.net> (raw)

Hi,

I believe I've found a bug in the routing cache (perhaps only for
IPv6) in Linux 3.0.4.  Since I'm uncertain about the diagnosis, I'm
posting here for guidance as to how to proceed.

In a nutshell, the problem is this: my (manually entered) routing
tables specify an explicit MTU/MSS for certain IPv6 routes, and after
upgrading from 2.6.38.7 to 3.0.4, I found that this MTU/MSS setting is
sometimes ignored.  More precisely, it seems that connecting to a host
through the route in question causes a routing cache entry to be
created, and it is when this cache entry expires that the route
forgets the MTU/MSS setting.  (It may also be relevant that in my case
the route belongs to a routing table other than "main".)

In more details: my routing tables contains the following:

vega david ~ $ sudo ip -6 rule show 
0:      from all lookup local 
32766:  from all lookup main 
40000:  from 2002::/16 lookup 2002 
40100:  from all lookup 2001 
vega david ~ $ sudo ip -6 route show table 2001
2002::/16 via ::192.88.99.1 dev tun6to4  metric 512  mtu 1466 advmss 1406
2000::/3 dev ppp0  metric 64  mtu 1466 advmss 1406

Initially, trying to get the route to 2001:41d0:1:a431::1 correctly
returns route with the mtu and advmss values I entered:

vega david ~ $ sudo ip route get 2001:41d0:1:a431::1
2001:41d0:1:a431::1 from :: via 2001:41d0:1:a431::1 dev ppp0  src 2001:7a8:78ae::1  metric 0 
    cache  mtu 1466 advmss 1406

Now I start an ssh connection to 2001:41d0:1:a431::1, close it
immediately, and fetch the route again:

vega david ~ $ sudo ip route get 2001:41d0:1:a431::1
2001:41d0:1:a431::1 from :: via 2001:41d0:1:a431::1 dev ppp0  src 2001:7a8:78ae::1  metric 0 
    cache  mtu 1466 rtt 37ms rttvar 37ms cwnd 10 advmss 1406

This is still correct.  But if I wait a few seconds (doing nothing)
for these data to expire, they seems to take the mtu and advmss values
away with them:

vega david ~ $ sudo ip route get 2001:41d0:1:a431::1
2001:41d0:1:a431::1 from :: via 2001:41d0:1:a431::1 dev ppp0  src 2001:7a8:78ae::1  metric 0 
    cache 

(note the empty "cache" line and the absence of mtu/advmss settings).

This is not just wrongly reported: the route really is wrong, and
opening a TCP connection at this stage will use a wrong MSS (and
freeze on large packets, because my Internet provider is worthless,
but that's another story).  At some point the route seems to return to
normal, but I couldn't figure out what causes this exactly.

In case it's uesful, my kernel config is on <URL:
http://www.madore.org/~david/.tmp/config-3.0.4-vega
 >.

Any thoughts?

-- 
     David A. Madore
   ( http://www.madore.org/~david/ )

             reply	other threads:[~2011-09-01 21:56 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2011-09-01 21:46 David Madore [this message]
2011-09-08 14:42 ` possible routing table cache bug for mtu/advmss attributes in Linux 3.0.4 Yan, Zheng 
2011-09-09 21:28   ` David Madore

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20110901214643.GA16989@aldebaran.gro-tsen.net \
    --to=david+ml@madore.org \
    --cc=netdev@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox