[quagga-dev 4158] Re: Bug in long delay networks

Paul Jakma paul at clubi.ie
Wed May 31 13:10:02 BST 2006

On Wed, 31 May 2006, Vincent Jardin wrote:

> Agree, I do prefer this option.

Well, its the logical conclusion of Philips proposal to remove the 
flood call in the maxage_remover. ;)

The only problem is, this morning it had crashed. And it crashes 
whenever a non-self-originated LSA changes (it seems). There appears 
to be an ospf_lsa_lock() missing somewhere:

2006/05/31 06:31:11 OSPF: RXmtL(1)--, NBR(, LSA[Type1,id(,ar(]
2006/05/31 06:31:11 OSPF: LSA: freed 0x5c39d0
2006/05/31 06:31:11 OSPF: LSA[Type1:]: data freed 0x5c5680
2006/05/31 06:31:11 OSPF: SPF: calculation timer scheduled
2006/05/31 06:31:11 OSPF: SPF: calculation timer delay = 200
2006/05/31 06:31:11 OSPF: LSA[Type1,id(,ar(]: Install router-LSA to Area
2006/05/31 06:31:11 OSPF: LSA[Type1:]: Install LSA 0x0x5bf370, MaxAge
OSPF: Received signal 11 at 1149053471 (si_addr 0x0); aborting...
Backtrace for 14 stack frames:

The trace is very consistent.

And interestingly the reporter on Bug #269 has just added more 
information that strongly suggests his crash must be the same 
problem.. (So it's an existing and apparently mostly latent bug, i.e. 
*not* specific to this patch).

Where is the missing lsa_lock() though?

Paul Jakma	paul at clubi.ie	paul at jakma.org	Key ID: 64A2FF6A
Grandpa Charnock's Law:
 	You never really learn to swear until you learn to drive.

 	[I thought it was when your kids learned to drive.  Ed.]

More information about the Quagga-dev mailing list