I changed the MTU's for a bond interface & associated network devices to support jumbo frames. Ran "system network restart" & got the following traces. The same command ran on 2 other identical systems with no problems. Hardware is dl585 with e1000 NIC's. kernel is 2.6.24-0.14.rc8 (x86_64). Call Trace: [<ffffffff8043d7da>] fib_rules_event+0x4f/0x128 [<ffffffff804a89fe>] notifier_call_chain+0x33/0x5b [<ffffffff802566dd>] __raw_notifier_call_chain+0x9/0xb [<ffffffff802566ee>] raw_notifier_call_chain+0xf/0x11 [<ffffffff8042fabe>] call_netdevice_notifiers+0x16/0x18 [<ffffffff8042fd09>] netdev_state_change+0x22/0x36 [<ffffffff8043a5f5>] __linkwatch_run_queue+0x171/0x19e [<ffffffff8043a64c>] linkwatch_event+0x2a/0x31 [<ffffffff8024e7ef>] run_workqueue+0x108/0x1d8 [<ffffffff8043a622>] linkwatch_event+0x0/0x31 [<ffffffff8024f450>] worker_thread+0x0/0xea [<ffffffff8024f52f>] worker_thread+0xdf/0xea [<ffffffff802524bb>] autoremove_wake_function+0x0/0x38 [<ffffffff80252395>] kthread+0x49/0x78 [<ffffffff8020d068>] child_rip+0xa/0x12 [<ffffffff8025234c>] kthread+0x0/0x78 [<ffffffff8020d05e>] child_rip+0x0/0x12 RTNL: assertion failed at net/ipv4/devinet.c (1054) Pid: 107, comm: events/0 Not tainted 2.6.24-0.14.rc8.el5rt #1 Call Trace: [<ffffffff8047cb79>] inetdev_event+0x5c/0x46b [<ffffffff8027daa9>] __rcu_read_unlock+0x5a/0x5c [<ffffffff804a89fe>] notifier_call_chain+0x33/0x5b [<ffffffff802566dd>] __raw_notifier_call_chain+0x9/0xb [<ffffffff802566ee>] raw_notifier_call_chain+0xf/0x11 [<ffffffff8042fabe>] call_netdevice_notifiers+0x16/0x18 [<ffffffff8042fd09>] netdev_state_change+0x22/0x36 [<ffffffff8043a5f5>] __linkwatch_run_queue+0x171/0x19e [<ffffffff8043a64c>] linkwatch_event+0x2a/0x31 [<ffffffff8024e7ef>] run_workqueue+0x108/0x1d8 [<ffffffff8043a622>] linkwatch_event+0x0/0x31 [<ffffffff8024f450>] worker_thread+0x0/0xea [<ffffffff8024f52f>] worker_thread+0xdf/0xea [<ffffffff802524bb>] autoremove_wake_function+0x0/0x38 [<ffffffff80252395>] kthread+0x49/0x78 [<ffffffff8020d068>] child_rip+0xa/0x12 [<ffffffff8025234c>] kthread+0x0/0x78 [<ffffffff8020d05e>] child_rip+0x0/0x12
Graham, Have you tried this on 2.6.24.3-rt3 (which is currently only available in brew)? Clark
Arnaldo, Have you seen this with bonding on the latest kernels? Clark
I had only one network setup with bonding, on Shak's machines, will try to setup it with rhelrt2 and 5.
Tested now with just one bonding slave on rhel-rt2/5, changing the MTU of the slave and of bond0 back and forth from/to 1500/9000, directly with ifconfig and thru 'service network reload', no problem at all with recent kernels. Was this observed with recent kernels? Perhaps this is a driver problem? rhelrt-2 has a bnx2 card and rhelrt-5 has a e1000 card.
Tried now with: [root@rhelrt-2 network-scripts]# uname -r 2.6.24-0.10.rc8.el5rt Couldn't reproduce with this kernel either.