ports/78302: net/zebra patch for ospfd lost database updates

[Dmitry Morozovsky]

>Number:         78302
>Category:       ports
>Synopsis:       net/zebra patch for ospfd lost database updates
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    freebsd-ports-bugs
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Wed Mar 02 14:50:14 GMT 2005
>Closed-Date:
>Last-Modified:
>Originator:     Dmitry Morozovsky
>Release:        FreeBSD 4-STABLE i386
>Organization:
Cronyx Plus LLC (RiNet ISP)
>Environment:
System: FreeBSD 4-STABLE 

>Description:

It seems to be a bug in zebra's LSA Refresh algorythm when time got backwards,
where refresh loop can go to virtually endless. We sometimes encounted this in
rather simple topology, but with many (3k+) LSAs

>How-To-Repeat:

Non-deterministic, but mostly working example:
run zebra's ospfd, slide time backwards. watch for adjacency goes down on other
OSPF router.

>Fix:

We're run zebra with this fix for more than half a year now. Everything seems
to be ok.

Index: Makefile
===================================================================
RCS file: /home/ncvs/ports/net/zebra/Makefile,v
retrieving revision 1.79
diff -u -r1.79 Makefile
--- Makefile	7 Feb 2005 12:36:46 -0000	1.79
+++ Makefile	2 Mar 2005 14:35:35 -0000
@@ -7,7 +7,7 @@
 
 PORTNAME=	zebra
 PORTVERSION=	0.94
-PORTREVISION=	2
+PORTREVISION=	3
 CATEGORIES=	net ipv6
 MASTER_SITES=	ftp://ftp.zebra.org/pub/zebra/ \
 		ftp://ftp.ripe.net/mirrors/sites/ftp.zebra.org/pub/zebra/ \
Index: files/patch-ospfd_ospf__lsa.c
===================================================================
RCS file: files/patch-ospfd_ospf__lsa.c
diff -N files/patch-ospfd_ospf__lsa.c
--- /dev/null	1 Jan 1970 00:00:00 -0000
+++ files/patch-ospfd_ospf__lsa.c	2 Mar 2005 14:35:35 -0000
@@ -0,0 +1,37 @@
+--- ospfd/ospf_lsa.c.orig	Mon Nov  8 11:47:25 2004
++++ ospfd/ospf_lsa.c	Tue Nov  9 03:15:53 2004
+@@ -3344,6 +3344,7 @@
+   listnode node;
+   struct ospf *ospf = THREAD_ARG (t);
+   int i;
++  time_t ct;
+   list lsa_to_refresh = list_new ();
+ 
+   if (IS_DEBUG_OSPF (lsa, LSA_REFRESH))
+@@ -3351,15 +3352,25 @@
+ 
+   
+   i = ospf->lsa_refresh_queue.index;
++
++  if ((ct = time(NULL)) < ospf->lsa_refresher_started) {
++    zlog_info ("LSA[Refresh]: ospf_lsa_refresh_walker(): HACK: fixing lsa_refresher_started: %d -> %d", ospf->lsa_refresher_started, ct);
++    ospf->lsa_refresher_started = ct;
++  }
+   
+   ospf->lsa_refresh_queue.index =
+     (ospf->lsa_refresh_queue.index +
+-     (time (NULL) - ospf->lsa_refresher_started) / OSPF_LSA_REFRESHER_GRANULARITY)
++     (ct - ospf->lsa_refresher_started) / OSPF_LSA_REFRESHER_GRANULARITY)
+     % OSPF_LSA_REFRESHER_SLOTS;
+ 
+   if (IS_DEBUG_OSPF (lsa, LSA_REFRESH))
+     zlog_info ("LSA[Refresh]: ospf_lsa_refresh_walker(): next index %d",
+ 	       ospf->lsa_refresh_queue.index);
++
++  if (ospf->lsa_refresh_queue.index >= OSPF_LSA_REFRESHER_SLOTS) {
++    zlog_info ("LSA[Refresh]: ospf_lsa_refresh_walker(): HACK: fixing index: %hu -> 0", ospf->lsa_refresh_queue.index);
++    ospf->lsa_refresh_queue.index = 0;
++  }
+ 
+   for (;i != ospf->lsa_refresh_queue.index;
+        i = (i + 1) % OSPF_LSA_REFRESHER_SLOTS)
>Release-Note:
>Audit-Trail:
>Unformatted: