Hi, I have been testing the carp link state patch from last week (which works perfectly, ifconfig shows correct state every time now).
ospfd however, sometime incorrectly think some interfaces are in state "invalid" instead of "down" on startup: Interface Address State HelloTimer Linkstate Uptime nc ac carp7 195.58.98.145/28 DOWN - backup 00:00:00 0 0 carp5 192.168.253.1/24 DOWN - backup 00:00:00 0 0 carp3 192.168.202.1/24 DOWN - backup 00:00:00 0 0 carp2 192.168.254.1/23 DOWN - backup 00:00:00 0 0 carp1 31.15.61.129/26 DOWN - invalid 00:00:00 0 0 carp0 92.33.0.202/30 DOWN - invalid 00:00:00 0 0 bnx0 192.168.200.5/24 OTHER 00:00:02 active 00:01:47 4 2 after restarting ospfd, it detects the correct link state: Interface Address State HelloTimer Linkstate Uptime nc ac carp7 195.58.98.145/28 DOWN - backup 00:00:00 0 0 carp5 192.168.253.1/24 DOWN - backup 00:00:00 0 0 carp3 192.168.202.1/24 DOWN - backup 00:00:00 0 0 carp2 192.168.254.1/23 DOWN - backup 00:00:00 0 0 carp1 31.15.61.129/26 DOWN - backup 00:00:00 0 0 carp0 92.33.0.202/30 DOWN - backup 00:00:00 0 0 bnx0 192.168.200.5/24 OTHER 00:00:00 active 00:01:29 4 2 I figure this is some kind of race when ospfd initializes at the same time as the kernel initializes all interfaces. >From the debug log it seems that ospfd misses some events: May 26 20:25:19 fw2 ospfd[380]: startup May 26 20:25:22 fw2 ospfd[5184]: interface carp3 up May 26 20:25:23 fw2 ospfd[5184]: interface carp5 up May 26 20:25:23 fw2 ospfd[5184]: interface carp7 up May 26 20:25:23 fw2 ospfd[5184]: interface carp2 up May 26 20:25:23 fw2 ospfd[5184]: interface carp0 up May 26 20:25:23 fw2 ospfd[5184]: interface carp0 down May 26 20:25:23 fw2 ospfd[5184]: interface carp7 down May 26 20:25:23 fw2 ospfd[5184]: interface carp2 down May 26 20:25:23 fw2 ospfd[5184]: interface carp5 down In this case, carp1 was left in the invalid state (but ifconfig shows backup state). carp3 also looks suspicious in the log (missing a down event), but is correctly reported as backup by ospfctl. >From what I can see the event handler is correctly registered before trying to start all interfaces in ospfe() so I really don't understand where the race is. /Johan Ymerson
