Now, let me try to understand this regression. The kernel log you sent shows the interface being brought up and down several times in quick succession: First time: [ 224.139579] tg3 0000:04:04.0: irq 30 for MSI/MSI-X [ 224.174436] ADDRCONF(NETDEV_UP): eth0: link is not ready [ 224.185365] tg3 0000:04:04.0: eth0: Link is up at 1000 Mbps, full duplex [ 224.185367] tg3 0000:04:04.0: eth0: Flow control is off for TX and off for RX [ 224.194074] ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready [ 224.226364] tg3 0000:04:04.0: PME# enabled Second time: [ 225.417581] tg3 0000:04:04.0: irq 30 for MSI/MSI-X [ 225.451422] ADDRCONF(NETDEV_UP): eth0: link is not ready [ 225.463365] tg3 0000:04:04.0: eth0: Link is up at 1000 Mbps, full duplex [ 225.463368] tg3 0000:04:04.0: eth0: Flow control is off for TX and off for RX Third time: [ 227.045107] tg3 0000:04:04.0: irq 30 for MSI/MSI-X [ 227.062399] ADDRCONF(NETDEV_UP): eth0: link is not ready [ 227.091050] tg3 0000:04:04.0: eth0: Link is up at 1000 Mbps, full duplex [ 227.091053] tg3 0000:04:04.0: eth0: Flow control is off for TX and off for RX [ 227.092177] ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready Could you test whether the same problem occurs if you do something similar after reloading the driver: rmmod tg3 modprobe tg3 ip link set eth0 up ; sleep 0.1 ; ip link set eth0 down ip link set eth0 up ; sleep 0.1 ; ip link set eth0 down ip link set eth0 up If that does reproduce the problem, could you also try the same with Linux 3.0 on this system? Ben. -- Ben Hutchings No political challenge can be met by shopping. - George Monbiot
Attachment:
signature.asc
Description: This is a digitally signed message part