Hello everybody. For a few months I've been experiencing this problem - it was a bit hard to track because it usually happens only during shutdown, when network interfaces go down, so I just didn't notice it. A kernel panic happens when one of the interfaces, provided by tg3 driver goes down. "ifdown eth2" is enough to cause it.
It doesn't matter if this interface was actively used or even if the link was up - I can unplug the cable, boot up (interface is configured to use dhcp, it will attempt to go up and fail), then execute "ifdown eth2" and system will crash. It's a bit hard to get the full message of the crash as this happens on the machine which I use for remote logging itself.. The best thing I can get right away are screenshots, however, some of information might be missing on them. It goes like this on shutdown (or "ifdown eth2", or "service network restart" etc): 1) interfaces are being brought down, at some point eth2 is being brought down 2) nothing happens for about 10 seconds, system appears to be hang 3) lots of lines with call traces appear and scroll through the screen. These are last lines which I captured in screenshot: http://img202.imageshack.us/img202/5459/20121225205828.png 4) about 10 second pause again 5) kernel panic happens, more lines scroll. Again, here are some of the last ones: http://img5.imageshack.us/img5/397/20121225205838.png 6) system hangs completely This happens on latest kernel-2.6.32-279.19.1.el6.x86_64. It also happened on 2.6.32-279.11.1.el6.x86_64 and 2.6.32-279.14.1.el6.x86_64. It didn't happen in SL6.2 with (official, not from elrepo) kmod-tg3-3.122 package installed which was present in 6.2-fastbugs repository. I found some information about tg3 crashes like this http://elrepo.org/bugs/view.php?id=315 or this http://bugs.centos.org/view.php?id=5428 but in either case 3.122 version of tg3 driver solved the problem. However, I'm already using 3.122 and still experience crash. The controller in question is Broadcom NetXtreme BCM5701, PCI-X version which is inserted into PCI-X slot of Supermicro X7SBE. There haven't been any hardware changes lately and it is working stable. I'm pretty sure that this bug has appeared somewhere along the 6.2->6.3 upgrade or in one of the 6.3 kernels. It's a bit hard to track because it appears simply as "hang during reboot or shutdown", which rarely happens for this system, but I'm sure that few months ago it rebooted and powered off just fine. This is interface used for internet connection. VLANs are not used. There exists sixxs-based IPv6 interface in system, configured to work over this interface. This problem doesn't happen with other (intel e1000e) network interfaces. $ cat /etc/sysconfig/network-scripts/ifcfg-eth2 DEVICE=eth2 BOOTPROTO=dhcp ONBOOT=yes TYPE=Ethernet HWADDR=00:02:A5:E7:0A:10 PEERDNS=no NOZEROCONF=yes $ ifconfig eth2 eth2 Link encap:Ethernet HWaddr 00:02:A5:E7:0A:10 inet addr:<skipped...> inet6 addr: fe80::202:a5ff:fee7:a10/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:130906804 errors:0 dropped:0 overruns:0 frame:0 TX packets:178575110 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:100 RX bytes:83971053482 (78.2 GiB) TX bytes:205754543966 (191.6 GiB) Interrupt:52 $ dmesg|grep '\(eth2\|tg3\)' tg3.c:v3.122 (December 7, 2011) tg3 0000:03:02.0: PCI INT A -> GSI 52 (level, low) -> IRQ 52 tg3 0000:03:02.0: eth2: Tigon3 [partno(253212-001) rev 0105] (PCIX:133MHz:64-bit) MAC address 00:02:a5:e7:0a:10 tg3 0000:03:02.0: eth2: attached PHY is 5701 (10/100/1000Base-T Ethernet) (WireSpeed[1], EEE[0]) tg3 0000:03:02.0: eth2: RXcsums[0] LinkChgREG[0] MIirq[0] ASF[0] TSOcap[0] tg3 0000:03:02.0: eth2: dma_rwctrl[76db000f] dma_mask[64-bit] ADDRCONF(NETDEV_UP): eth2: link is not ready tg3 0000:03:02.0: eth2: Link is up at 100 Mbps, full duplex tg3 0000:03:02.0: eth2: Flow control is on for TX and on for RX ADDRCONF(NETDEV_CHANGE): eth2: link becomes ready Does anyone know some solution or workaround? I'm fine with installing other version of this driver from kmod (if I knew where to get better version), but not very comfortable with using kernel-3.5/3.6/3.7 etc from elrepo. -- Vladimir