On 11/19/18 7:16 PM, David Miller wrote: > From: Xin Long <lucien....@gmail.com> > Date: Thu, 15 Nov 2018 16:23:38 +0900 > >> The attachment is the ip6_dst.sh with IPVS. >> >> # sh ip6_dst.sh > > Maybe a selftests candidate? >
That script was not a reliable reproducer for me. I created a much simpler one that shows the problem every time - and proves the change. It needs some adjustments to work as a selftest. Specifically, it needs to check dmesg for net_device reference counts after the rcu delay. I'll add to the to-do list.
