PPPoE MTU / MRU - PPP

This is a discussion on PPPoE MTU / MRU - PPP ; I had my openbsd system set up as a firewall / router for my home network and things were working without any problems for many months. My ISP provides a PPPoE connection over DSL. I was using the stock userland ...

+ Reply to Thread
Results 1 to 4 of 4

Thread: PPPoE MTU / MRU

  1. PPPoE MTU / MRU

    I had my openbsd system set up as a firewall / router for my home
    network and things were working without any problems for many months.
    My ISP provides a PPPoE connection over DSL. I was using the stock
    userland PPP software that ships with OpenBSD, which I believe is widely
    used on many platforms.

    Recently (during the last 4 days), I had a lot of network problems that
    I eventually solved by dropping the MTU in my PPP configuration from
    1492 to 1488. I could not even get full packets with MTU 1492 out to my
    ISP's nearest router all of a sudden...does this strike anyone else as
    odd? I called my ISP and they (who generally seem fairly competent) say
    they made no configuration changes.

    Also, this got me to realizing that I had no idea why I've even got a
    max MRU in my ppp config specified as 1492 (aside from the obvious: I
    copied it from someone else's config file). Without knowing much of
    anything about networking in general, and less about PPP in specific,
    I'm in the dark as to why you'd ever specify a max MRU for a PPP
    connection. My ethernet card only has an MTU setting, AFAIK, so why is
    there an MRU for PPP? It seems to me that as a person on the receiving
    end of a packet transmission, I'll accept whatever size packet that
    comes down the pipe. I would have expected the MTU (thereby implying
    MSS?) to be enough to specify to anyone on the other end of a connection
    what my pipeline size is...can anyone explain this to me? Why do I need
    to set a separate MRU setting for PPP connections, and should I change
    it to 1488 as well now that I've had to change my MTU setting?

  2. Re: PPPoE MTU / MRU

    Paul Galbraith writes:
    > Recently (during the last 4 days), I had a lot of network problems
    > that I eventually solved by dropping the MTU in my PPP configuration
    > from 1492 to 1488. I could not even get full packets with MTU 1492
    > out to my ISP's nearest router all of a sudden...does this strike
    > anyone else as odd? I called my ISP and they (who generally seem
    > fairly competent) say they made no configuration changes.


    Yes, it's odd. Something is broken along the path, though it's hard
    to tell from the information given what that might be.

    If it started suddenly, then I'm not sure I'd completely believe the
    assertion by the ISP. One thing to realize is that there are usually
    several parties involved with a DSL connection; at least two and
    perhaps as many as four: a local telco, a DSL provider, an internal
    ATM network, and the ISP itself. It's possible that the ISP didn't
    cause the damage, but one of the others did. Someone with access to
    those networks might be able to trace the problem.

    It's the nature of PPPoE that it's really unable to cope with (or even
    know about) path problems like that.

    > Also, this got me to realizing that I had no idea why I've even got a
    > max MRU in my ppp config specified as 1492 (aside from the obvious: I
    > copied it from someone else's config file). Without knowing much of


    PPPoE (RFC 2516) imposes 8 octets of overhead on each packet. Since
    the user payload portion of Ethernet is limited to 1500 octets, that
    leaves 1492 available for the network layer (such as IP).

    For what it's worth, PPPoE's bizarre MTU/MRU tends to cause all sorts
    of problems with various sites due to problems with Path MTU
    Discovery. There's really no good fix for this. There are many web
    sites that lament the problem, though.

    > anything about networking in general, and less about PPP in specific,
    > I'm in the dark as to why you'd ever specify a max MRU for a PPP
    > connection. My ethernet card only has an MTU setting, AFAIK, so why
    > is there an MRU for PPP? It seems to me that as a person on the


    PPP is symmetric: the peers agree on all of the parameters in both
    directions, so there's both an MRU and an MTU. It's just how it's
    done. (See RFC 1661.)

    > receiving end of a packet transmission, I'll accept whatever size
    > packet that comes down the pipe. I would have expected the MTU
    > (thereby implying MSS?) to be enough to specify to anyone on the other
    > end of a connection what my pipeline size is...can anyone explain this
    > to me? Why do I need to set a separate MRU setting for PPP
    > connections, and should I change it to 1488 as well now that I've had
    > to change my MTU setting?


    As a PPP person, I find Ethernet's conflation of send and receive to
    be confusing. ;-}

    MTU says "this is the largest packet I know how to send through this
    link." MRU says "this is the largest packet I am willing to receive."

    Of course, *good* implementations actually are able to receive at
    least a few octets over the specified MRU, in order to deal with peers
    that aren't so good at arithmetic.

    --
    James Carlson, IP Systems Group
    Sun Microsystems / 1 Network Drive 71.234W Vox +1 781 442 2084
    MS UBUR02-212 / Burlington MA 01803-2757 42.497N Fax +1 781 442 1677

  3. Re: PPPoE MTU / MRU

    James Carlson wrote:

    > For what it's worth, PPPoE's bizarre MTU/MRU tends to cause all sorts
    > of problems with various sites due to problems with Path MTU
    > Discovery. There's really no good fix for this. There are many web
    > sites that lament the problem, though.


    Hmm. I thought it was the web sites themselves that were causing the
    problem - by blocking ICMP type 3 code 4 messages, fragmentation needed
    but DF bit set, and so preventing PMTU Discovery. Even then it seems
    to me that the problem should most often (only?) occur for masqueraded
    hosts on a LAN behind a PPPoE connection host.

    In this case I agreed with your assessment that something changed at
    the ISP or the other associated entities you mentioned. And I've seen
    a number of other posts that said a similar problem was resolved by
    reducing the PPPoE MTU to some value below 1492. The particular value
    seems to vary.

    --
    Clifford Kite Email: "echo xvgr_yvahk-ccc@ri1.arg|rot13"
    PPP-Q&A links, downloads: http://ckite.no-ip.net/
    /* The generation of random numbers is too important to be left
    to chance. */

  4. Re: PPPoE MTU / MRU

    Clifford Kite writes:
    > James Carlson wrote:
    >
    > > For what it's worth, PPPoE's bizarre MTU/MRU tends to cause all sorts
    > > of problems with various sites due to problems with Path MTU
    > > Discovery. There's really no good fix for this. There are many web
    > > sites that lament the problem, though.

    >
    > Hmm. I thought it was the web sites themselves that were causing the
    > problem - by blocking ICMP type 3 code 4 messages, fragmentation needed
    > but DF bit set, and so preventing PMTU Discovery.


    Right. It's essentially the result of a culture clash between the
    folks who designed Path MTU Discovery and the ones who implement
    "security." Regardless of the reason, though, the effect is that
    PMTUD doesn't always work very well.

    > Even then it seems
    > to me that the problem should most often (only?) occur for masqueraded
    > hosts on a LAN behind a PPPoE connection host.


    I've also heard reports of trouble reaching some sites directly over
    such a link (not sure how much stock I place in them, though). You're
    right that either masquerading or just plain old routing exacerbates
    the problem.

    --
    James Carlson, IP Systems Group
    Sun Microsystems / 1 Network Drive 71.234W Vox +1 781 442 2084
    MS UBUR02-212 / Burlington MA 01803-2757 42.497N Fax +1 781 442 1677

+ Reply to Thread