headers
Milan Kocian | 1 Dec 2007 10:07
Picon

[quagga-dev 5151] Re: Shadowing Daemon (implementing callback from zclient.h)

On Fri, Nov 30, 2007 at 01:35:30PM -0500, Eric Keller wrote:
>
>>> ip route change 192.169.0.0/16 via 192.168.100.2
>>> Zebra doesn't print anything
>>>
>>> ip route change 192.169.0.0/16 dev nf2c0
>>> Zebra doesn't print anything
>>>
>>> ip route change 192.169.0.0/16 dev nf2c0 via 192.168.101.2
>>> Zebra doesn't print anything
>>
>> What does the Linux 'ip monitor' say?
>>
>
> 'ip monitor' doesn't print anything for changes.  For example changing 
> 192.169.0.0/16 from using dev nf2c1 to using dev nf2c0 (which are 
> physically different devices). 
> With both zebra and my shadowing daemon running, I do the ip route change.  
> Then, if I keep zebra running, and quit my shadowing daemon and then 
> restart it, the route that get distributed to the shadowing daemon is the 
> old route (192.169.0.0/16 using nf2c1).
>
> If I quit out of both, then restart both.  Then the shadowing daemon gets 
> sent the latest entry (192.169.0.0/16 using nf2c0).
>
hi,

When you use linux kernel, it is known bug. When you replace route via 'ip r r'
kernel doesn't send netlink message. Fixed in 2.6.23.
(Continue reading)

Permalink | Reply | 0 comments |
headers
Daniel Ng | 3 Dec 2007 00:44
Picon

[quagga-dev 5152] Re: [PATCH] RIPv2 Redistributing Static Routes

Paul Jakma <Paul.Jakma <at> Sun.COM> writes:
> > +              case ZEBRA_ROUTE_RIP:
> > +              case ZEBRA_ROUTE_STATIC:
> > +              {
> > +                if (rinfo->ifindex == ifc->ifp->ifindex)
> > +                  continue;
> > +                break;
> > +              }
> 
> Wasn't it the idea that this would be the default case?
> 

Paul, if we make this the the default case then we will miss out executing 
the 'route-map' code later on in the function. This 'route-map' code also 
deals with the '(rinfo->ifindex == ifc->ifp->ifindex)' case. 

Do you agree?
Permalink | Reply | 2 comments |
headers
Christian Hammers | 3 Dec 2007 09:21
Picon
Favicon

[quagga-dev 5153] Re: Bug#454096: quagga: ospf6d crashes w/`orig->retrans_count >= 0' assertion failure

Hello

The following is forwarded from the Debian Bug Tracking System. Any ideas?

bye,

-christian

On 2007-12-02 Mark Kamichoff wrote:
> Package: quagga
> Version: 0.99.9-2
> Severity: normal
> 
> 
> Ever since upgrading to quagga 0.99.9-2, the ospf6d daemon crashes every
> few days, and I see the following in the logs each time:
> 
> 2007/12/02 19:33:30 OSPF6: Assertion `orig->retrans_count >= 0' failed in
> file ospf6_flood.c, line 203, function ospf6_decrement_retrans_count
> 2007/12/02 19:33:30 OSPF6: Backtrace for 15 stack frames: 2007/12/02
> 19:33:30 OSPF6: [bt 0] /usr/lib/libzebra.so.0(zlog_backtrace+0x1f)
> [0xb7edcecb] 2007/12/02 19:33:30 OSPF6: [bt
> 1] /usr/lib/libzebra.so.0(_zlog_assert_failed+0x99) [0xb7edd028]
> 2007/12/02 19:33:30 OSPF6: [bt
> 2] /usr/lib/quagga/ospf6d(ospf6_decrement_retrans_count+0x51) [0x805e751]
> 2007/12/02 19:33:30 OSPF6: [bt
> 3] /usr/lib/quagga/ospf6d(ospf6_flood_clear_interface+0x5f) [0x805e7b5]
> 2007/12/02 19:33:30 OSPF6: [bt
> 4] /usr/lib/quagga/ospf6d(ospf6_flood_clear_area+0x22) [0x805e83e]
> 2007/12/02 19:33:30 OSPF6: [bt
(Continue reading)

Permalink | Reply | 0 comments |
headers
Paul Jakma | 6 Dec 2007 08:31
Picon

[quagga-dev 5154] Re: Shadowing Daemon (implementing callback from zclient.h)

On Fri, 30 Nov 2007, Eric Keller wrote:

> 'ip monitor' doesn't print anything for changes.

So Linux does not notify anyone if a route is changed? Sounds like a 
Linux bug.

regards,
--

-- 
Paul Jakma,
Solaris Networking                       Sun Microsystems, Scotland
http://opensolaris.org/os/project/quagga tel: EMEA x73150 / +44 15066 73150
Permalink | Reply | 1 comment |
headers
Eric Keller | 6 Dec 2007 13:57
Picon
Favicon

[quagga-dev 5155] Re: Shadowing Daemon (implementing callback from zclient.h)

Paul Jakma wrote:
> On Fri, 30 Nov 2007, Eric Keller wrote:
>
>> 'ip monitor' doesn't print anything for changes.
>
> So Linux does not notify anyone if a route is changed? Sounds like a 
> Linux bug.
>
> regards,

I agree.  Milan pointed to the report that it is fixed in 2.6.23 
(thanks, Milan).
Thanks again,
Eric
Permalink | Reply | 0 comments |
headers
Michael Larson | 21 Dec 2007 21:18
Favicon

[quagga-dev 5156] interface static routes and link state awareness

Hello all,

My first time posting to this list.

I'm looking into static routes and link state awareness (physical link 
up/down, or IFF_RUNNING).

A recent bug fix (bug 418) was applied that allowed Quagga to remove 
static routes with nexthops from the kernel rib when IFF_RUNNING changed 
from true->false. And correspondingly, the routes would be reinserted 
when IFF_RUNNING changed from false->true.

However, this fix only applied to static routes with nexthops and not to 
static interface routes. Static interface routes still remain active in 
the kernel rib when IFF_RUNNING changed from true->false.

I've dug through the source a bit and found a fix that works for static 
interface routes (the patch is inserted at the end of this email). The 
patch involved changing the static route test from testing against the 
IFF_UP flag to the IFF_RUNNING flag. Since I have only a fleeting 
familiarity with Quagga at this point--I'd appreciate it if others could 
weigh in on this patch and let me know if this is a bad idea (or a good 
idea). Thanks all!

Mike Larson

 >>>>>>>>>>>>>>>>Patch below>>>>>>>>>>>>>>>>>>>>>

  --- zebra_rib.c 2007-12-21 12:01:12.000000000 -0800
  +++ zebra_rib_fix.c     2007-12-21 12:02:41.000000000 -0800
(Continue reading)

Permalink | Reply | 0 comments |
headers
paul | 22 Dec 2007 17:47
Picon

[quagga-dev 5157] Re: bgpd crash fixes (for review)

Not received any comment on this.

As they're /nearly/ all fairly obvious fixes, I'm going to commit 
this so as to try get wider testing.

I have since tested pushing ORF prefix-lists via Route-Refresh, and 
that appears to work - though there seem to be other problems with 
Quagga applying ORF filters.

The extended-length check seems fine.

On Mon, 19 Nov 2007, Paul Jakma wrote:

> Hi,
>
> I've been sitting on the attached fixes for a while, hoping that the reporter 
> would be able to verify whether it fixed the problems. Unfortunately that 
> hasn't possible, so the patch needs review here.
>
> - Fixes for a problem in the AS4 updates, where as4_path could be
>  dereferenced even if aspath_parse failed to set as4_path (should be
>  obviously correct)
>
> - incorrect assert, testing that AS4_AGGREGATOR implied we'd received
>  an aggregator already (and hence attr->extra was assumed valid).
>  That need not be the case. (fix again is reasonably obvious)
>
> - Failure to properly sanity check extended length attribute header
>  for minimum length
>  (this bit of the patch looks incomplete - i'll need to go recheck
(Continue reading)

Permalink | Reply | 1 comment |
headers
paul | 22 Dec 2007 18:05
Picon

[quagga-dev 5158] Re: [PATCH] RIPv2 Redistributing Static Routes

On Sun, 2 Dec 2007, Daniel Ng wrote:

> Paul Jakma <Paul.Jakma <at> Sun.COM> writes:
>>> +              case ZEBRA_ROUTE_RIP:
>>> +              case ZEBRA_ROUTE_STATIC:
>>> +              {
>>> +                if (rinfo->ifindex == ifc->ifp->ifindex)
>>> +                  continue;
>>> +                break;
>>> +              }
>>
>> Wasn't it the idea that this would be the default case?
>>
>
> Paul, if we make this the the default case then we will miss out executing
> the 'route-map' code later on in the function. This 'route-map' code also
> deals with the '(rinfo->ifindex == ifc->ifp->ifindex)' case.
>
> Do you agree?

I'm confused then. Why isn't that later code sufficient for 
ZEBRA_ROUTE_STATIC then?

Why should ripd be treating ZEBRA_ROUTE_STATIC differently to other 
ZEBRA_ROUTE types (other than ZEBRA_ROUTE_RIP and 
ZEBRA_ROUTE_CONNECT)?

Your initial argument was that we could create loops because we 
weren't applying split-horizon to routes generally, using statics as 
an example. Why shouldn't that argument apply to BGP, OSPF, etc..
(Continue reading)

Permalink | Reply | 1 comment |
headers
paul | 24 Dec 2007 09:20
Picon

[quagga-dev 5159] Re: bgpd crash fixes (for review)

On Sat, 22 Dec 2007, paul <at> clubi.ie wrote:

> I have since tested pushing ORF prefix-lists via Route-Refresh, and 
> that appears to work - though there seem to be other problems with 
> Quagga applying ORF filters.

This may be because I've broken the REMOVE_ALL ORF with the security 
fix (turns out it /is/ valid to have less than 7 bytes of ORF ;) ). 
Will fix over the holidays sometime.

regards,
--

-- 
Paul Jakma	paul <at> clubi.ie	paul <at> jakma.org	Key ID: 64A2FF6A
Fortune:
Oliver's Law:
 	Experience is something you don't get until just after you need it.
Permalink | Reply | 0 comments |
headers
ChengHui Hong | 27 Dec 2007 09:41
Favicon

[quagga-dev 5160] multiple UPDATE with a single BGP update

Hi,

   I noticed that sometimes cisco routers send out bgp updates with packet
containing multiple update messages, within the same TCP data segment, there
are 2 bgp marker field for 2 diff bgp updates, sometimes it contains 1 bgp
UPDATE and 1 KEEPALIVE within the same packet, is this kind of encapsulation
supported by quagga? , and is this mentioned in any RFCs?

   T.I.A.
Permalink | Reply | 2 comments |

Gmane