Hello Ondrej In the thread "BIRD <-> Quagga Compatiblity" (started in May, 2007) I have already posted some informations. If you haven't access to the old thread, I could post the mails/infos again. Is possible that Quagga does something wrong? Best regards Elmar -----Original Message----- From: Ondrej Zajicek [mailto:santiago@crfreenet.org] Sent: Mittwoch, 4. Februar 2009 23:43 To: Vonlanthen, Elmar Cc: bird-users@network.cz Subject: Re: Bird to quagga interoperation (WARNING!!! PGP with incorrect signature) On Wed, Feb 04, 2009 at 11:30:52AM +0100, Vonlanthen, Elmar wrote:
Hello
We still have some hosts with the quagga routing daemon in a bird environment.
With the bird releases 1.0.11 and 1.0.12 it is not working. Bird gets the routes, but quagga doesn't.
Can you provide a better description of your network setting? The current code seems to me being fully RFC-compliant. I will look at it. The LSA generating code needs some review, but this is an issue i don't know about. -- Elen sila lumenn' omentielvo Ondrej 'SanTiago' Zajicek (email: santiago@crfreenet.org) OpenPGP encrypted e-mails preferred (KeyID 0x11DEADC3, wwwkeys.pgp.net) "To err is human -- to blame it on a computer is even more so."
On Thu, Feb 05, 2009 at 10:40:20AM +0100, Vonlanthen, Elmar wrote:
Hello Ondrej
In the thread "BIRD <-> Quagga Compatiblity" (started in May, 2007) I have already posted some informations. If you haven't access to the old thread, I could post the mails/infos again.
Thank you for a hint, i found that thread.
Is possible that Quagga does something wrong?
Perhaps Quagga has a different interpretation of some terms in OSPF RFC. There are Linux networking configurations (for example ptp network addresses) that does not fit well to assumptions used in OSPF RFC. -- Elen sila lumenn' omentielvo Ondrej 'SanTiago' Zajicek (email: santiago@crfreenet.org) OpenPGP encrypted e-mails preferred (KeyID 0x11DEADC3, wwwkeys.pgp.net) "To err is human -- to blame it on a computer is even more so."
In the thread "BIRD <-> Quagga Compatiblity" (started in May, 2007) I have already posted some informations. If you haven't access to the old thread, I could post the mails/infos again.
Thank you for a hint, i found that thread.
Is possible that Quagga does something wrong?
Perhaps Quagga has a different interpretation of some terms in OSPF RFC. There are Linux networking configurations (for example ptp network addresses) that does not fit well to assumptions used in OSPF RFC.
I have no problem to apply this little patch for every new release. But what is the difference between these two lines? 1.)ln->data = ifa->iface->index; 2.)ln->data = ipa_to_u32(ifa->iface->addr->ip); Could there be any problem (perhaps in a BIRD<->BIRD setup) if "ifa->iface->index" is never used in topology.c? Thanks. Best regards Elmi
On Mon, Feb 09, 2009 at 02:27:20PM +0100, Vonlanthen, Elmar wrote:
In the thread "BIRD <-> Quagga Compatiblity" (started in May, 2007) I have already posted some informations. If you haven't access to the old thread, I could post the mails/infos again.
Thank you for a hint, i found that thread.
Is possible that Quagga does something wrong?
Perhaps Quagga has a different interpretation of some terms in OSPF RFC. There are Linux networking configurations (for example ptp network addresses) that does not fit well to assumptions used in OSPF RFC.
I have no problem to apply this little patch for every new release. But what is the difference between these two lines?
1.)ln->data = ifa->iface->index;
2.)ln->data = ipa_to_u32(ifa->iface->addr->ip);
RFC expects that network link has either 'classical' IP addresses from one subnet or are 'unnumbered'. The first line is the rule for LSA data in a case of 'unnumbered' interface and second line is the rule for 'classical' IP subnet interface. Linux PtP addresses does not fit well to any of these cases and Bird handles them as 'unnumbered'. Perhaps Quagga handles them in a different manner and do some consistency checking. So i would like do some research about this issue and probably communicate this with Quagga developers to have consistent handling of Linux-specific network setting before merging some changes to it. -- Elen sila lumenn' omentielvo Ondrej 'SanTiago' Zajicek (email: santiago@crfreenet.org) OpenPGP encrypted e-mails preferred (KeyID 0x11DEADC3, wwwkeys.pgp.net) "To err is human -- to blame it on a computer is even more so."
On Mon, Feb 09, 2009 at 02:27:20PM +0100, Vonlanthen, Elmar wrote:
In the thread "BIRD <-> Quagga Compatiblity" (started in May, 2007) I have already posted some informations. If you haven't access to the old thread, I could post the mails/infos again.
Thank you for a hint, i found that thread.
Is possible that Quagga does something wrong?
Perhaps Quagga has a different interpretation of some terms in OSPF RFC. There are Linux networking configurations (for example ptp network addresses) that does not fit well to assumptions used in OSPF RFC.
I have no problem to apply this little patch for every new release. But what is the difference between these two lines?
1.)ln->data = ifa->iface->index;
2.)ln->data = ipa_to_u32(ifa->iface->addr->ip);
RFC expects that network link has either 'classical' IP addresses from one subnet or are 'unnumbered'. The first line is the rule for LSA data in a case of 'unnumbered' interface and second line is the rule for 'classical' IP subnet interface. Linux PtP addresses does not fit well to any of these cases and Bird handles them as 'unnumbered'. Perhaps Quagga handles them in a different manner and do some consistency checking.
So i would like do some research about this issue and probably communicate this with Quagga developers to have consistent handling of Linux-specific network setting before merging some changes to it.
Ok, many thanks in advance for your research. Best regards Elmar
participants (2)
-
Ondrej Zajicek -
Vonlanthen, Elmar