probable bug in pktgen

While testing my own pktgen code I noticed that I can cause spectacular
crashes by setting the 'multiskb' to > 1 and using VLAN interfaces.  I
believe this is because the virtual device changes the skb->dev pointer
when it xmits, and then pktgen tries to send exactly this pkt again,
even though the skb->dev is now hosed up.

Unfortunately, my patch will not do the kernel pktgen much good, but
next time someone is hacking there, maybe they can fix it :)

Ben

-- 
Ben Greear [off-list ref]
Candela Technologies Inc  http://www.candelatech.com