Hi Soren,
From your description, the hardware did receive packets. But tcpdump
showed nothing. Could you run the simple dtrace script in the
attachment? It prints values if e1000g passes the packets to the upper
layer. Once you run into the scenario, ping through e1000g0 interface
and run "dtrace -s e1000g_receive.d" in parallel.
Regards,
Miles
soren wrote:
e1000g maintains hardware statistics. You can get the
receiving packet
count by running "kstat -m e1000g -i 0 1 |grep
ipackets64". If the value
bumps as you ping the link partner, it's likely to
indicate the hardware
receives the ping packets. Otherwise the interface
just doesn't receive
anything. If the value does bump, could you run
wireshark or snoop on
the problematic machine to see if the outputs contain
any receiving packets?
I did a reboot last night, but this morning it was wedged again. Same symptoms:
the e1000g0 interface isn't receiving packets, even though it's sending packets.
The ipackets64 and opackets64 counters are incrementing, indicating that the interface is
sending and receiving packets. 'pfexec tcpdump -i e1000g0' doesn't print anything,
although when it quit it reports "1 packets captured; 533 packets received by
filter; 0 packets dropped".
A couple more weird things I noticed:
- Before rebooting, I was seeing the counters increment even when I wasn't
sending traffic to that IP. There was traffic on the network and tcpdump wasn't
running, but the counters were incrementing. After rebooting when things
started working properly the counters wouldn't increment unless they received
traffic intended for this machine.
- tcpdump is reporting that this is a 'link-type EN10MB' even though the router
lights indicate a gigabit link. Then again that's also what it says when I
reboot, so maybe that's just an imprecise reporting bug from tcpdump. Probably
nothing.
#!/usr/sbin/dtrace -Cs
#pragma D option flowindent
fbt:e1000g:e1000g_receive:return
/arg1 != NULL/
{
trace(arg1);
stack();
}
_______________________________________________
networking-discuss mailing list
[email protected]