[Dapper SRU] Assertion failure in OSPF

[Tore Anderson]

Binary package hint: quagga

The Dapper package of Quagga contains a critical bug that makes ospfd not start properly, rendering it completely unusable (and depending on the configuration, might ensure the machine never gets online). The Breezy version of Quagga did not suffer from this problem.

More detailed discussion, including a patch that later made it into the upstream sources, is available at <http://bugzilla.quagga.net/show_bug.cgi?id=229>.

I feel an update is warranted.

Tore.

[PrasannaL (prasanna79)]

This is a critical bug. Because of the race condition it makes using ospf with static routes unusable. Please consider a release.

The corresponding patch is available at
http://bugzilla.quagga.net/attachment.cgi?id=73&action=view

Thanks,
Prasanna.

[Brian Murray (brian-murray)]

Did you want to see this backported to Dapper or does it still exist in Feisty?

[PrasannaL (prasanna79)]

Backported to Dapper. I believe it was fixed in upstream before Edgy/Feisty.

Thanks for the quick response!

[PrasannaL (prasanna79)]

I can confirm that the patch applies cleanly to the dapper package (requires some --strip), and fixes the problem.

[John Dong (jdong)]

SRU-candidate patch available, please do bug fixes via SRU and not backports.

[Emmet Hikory (persia)]

Thank for for the patch, and interest in seeing it applied to Dapper. Please review https://wiki.ubuntu.com/StableReleaseUpdates for the procedure for requesting and deploying a stable release update. I've unsubscribed ubuntu-universe-sponsors, as 1) quagga is in main, and 2) there is no debdiff candidate revision for SRU review. Once a candidate revision including the requested patch is prepared, please subscribe ubuntu-main-sponsors to request upload.

[Martin Pitt (pitti)]

Patch available from upstream. There needs to be some research in the upstream bug tracker/svn about the history and feedback about this patch, then we should consider it for the dapper point release.

[Soren Hansen (soren)]

On Mon, Jul 23, 2007 at 01:37:33PM -0000, Martin Pitt wrote:
> Patch available from upstream. There needs to be some research in the
> upstream bug tracker/svn about the history and feedback about this
> patch, then we should consider it for the dapper point release.

I've gone through the patch and the current version of quagga and I can
at least say that no part of the patch has been reverted or even changed
at all. That's a pretty good start :)

--
Soren Hansen
Ubuntu Server Team
http://www.ubuntu.com/

[Henrik Nilsen Omma (henrik)]

Setting milestone to 6.06.2.

[Mathias Gug (mathiaz)]

I've attached a debdiff with the patch mentioned above.

I haven't be able to test that it fixes the bug as I don't have the proper setup to do this.

[Martin Pitt (pitti)]

I sponsored the upload with the diff, I just improved the changelog a bit:

 quagga (0.99.2-1ubuntu3.2) dapper-proposed; urgency=low
 .
   * debian/patches/92_router_id_not_timer_driven.dpatch:
     - Fix OSPF assertion on startup when using static routes. See
       http://bugzilla.quagga.net/show_bug.cgi?id=229 for details.
     - Patch taken from upstream CVS:
       http://bugzilla.quagga.net/attachment.cgi?id=73&action=view
     - LP #48848.

This needs verification now. Can anyone on dapper please install quagga and check that it still works and fixes that bug? Thank you!

[Martin Pitt (pitti)]

This still needs testing. Can you please give the -proposed packages a whirl, so that we can move them to -updates soon? Thanks!

[PrasannaL (prasanna79)]

Sorry, I missed this earlier.

The newest Ubuntu release seems to be quagga (0.99.2-1ubuntu3.3) . The patch is still needed, but the changelog needs to be modified.

[Mathias Gug (mathiaz)]

I've attached a patch against the latest version of quagga in dapper. However it still needs to be tested as I don't have the setup to test it.

[Daniel Holbach (dholbach)]

Shouldn't this be tested in dapper-proposed then?

[Daniel Holbach (dholbach)]

Is there something left for ubuntu-main-sponsors to do?

[Daniel Holbach (dholbach)]

Unsubscribing ubuntu-main-sponsors for now.

[Martin Pitt (pitti)]

Argh, sorry for the delay, ubuntu-sru was not subscribed. I subscribed it again, and uploaded Mathias' patch. 3.4 is in dapper-proposed now with the patch.

Can anyone please test this and give some instructions how to reproduce this? Thank you!

[Henrik Nilsen Omma (henrik)]

Brian writes:
"I gave this one a go also but was unable to setup OSPF in quagga altogether. To be certain this wasn't the bug I also tried it in Feisty
and with the dapper-proposed package but still wasn't able to setup OSPF. Without a test case I am uncertain what more we can do here."

[Martin Pitt (pitti)]

We need somebody to confirm that the version in dapper-proposed still works, and preferably even to test that this bug is fixed. This blocks the release of Ubuntu 6.06.2.

[Brian Murray (brian-murray)]

I tried using some config files I found at http://forums.suselinuxsupport.de/index.php?s=2200a0390e15e515869f19eb15b2d9c2&showtopic=40338&pid=179531&st=0&#entry179531
which seem to enable ospf. However, ospfd started up fine (the daemon was running and 'show ip ospf' returned information) for me using that config and package version 0.99.2-1ubuntu3.3.

[Mathias Gug (mathiaz)]

On Thu, Jan 17, 2008 at 01:25:18AM -0000, Brian Murray wrote:
> I tried using some config files I found at http://forums.suselinuxsupport.de/index.php?s=2200a0390e15e515869f19eb15b2d9c2&showtopic=40338&pid=179531&st=0&#entry179531
> which seem to enable ospf. However, ospfd started up fine (the daemon was running and 'show ip ospf' returned information) for me using that config and package version 0.99.2-1ubuntu3.3.

IIRC the bug could only be triggered under some specific configuration
(involving multiple servers exporting static routes). Reading the
upstream bug is a good starting point to figure out the configuration
that would trigger the bug.

--
Mathias

[Martin Pitt (pitti)]

Brian, Mathias, thanks for initial testing. I am not too worried about the specific patch, since in comment 4 confirmed that the patch fixes the problem. I'm more worried about toolchain changes and thus breaking the .debs in some way, etc.

[Martin Pitt (pitti)]

I consider this sufficiently verified.

[Martin Pitt (pitti)]

Copied to -updates