On 10/01/2017 15:36, Mason wrote: > I'm using kernel v4.9 on a [new] dev board. > I built a small kernel + rootfs with buildroot 2016.11.1 > eth0 is driven by drivers/net/ethernet/aurora/nb8800.c > > After booting, I just run udhcpc (busybox version) > Then I set the link down, and the kernel hangs.
So far, I've been unable to reproduce this issue on an older smp8758 dev board. # while true > do > ip link set dev eth0 down > sleep 3 > ip link set dev eth0 up > sleep 3 > let "++XX" > echo $XX > done The output is strange nonetheless: 1 2 3 4 5 [ 179.002757] nb8800 26000.ethernet eth0: Link is Up - 100Mbps/Full - flow control rx/tx 6 7 [ 192.538648] nb8800 26000.ethernet eth0: Link is Up - 1Gbps/Full - flow control rx/tx 8 9 10 11 12 [ 222.284936] nb8800 26000.ethernet eth0: Link is Up - 100Mbps/Full - flow control rx/tx 13 14 ^C # [ 235.800851] nb8800 26000.ethernet eth0: Link is Up - 1Gbps/Full - flow control rx/tx A) I would expect a "Link is Up" message for every "link up" command. I get only 4 messages for 14 commands. Is that expected? B) Twice, the PHY was only able to negotiate 100 Mbps, although the peer supports 1 Gbps. What could cause this? Regards.
