Can anyone explain to me why MLD Reports are sent to the Software Forwarding CPU Queue on a Catalyst 2960 or 2960X switch?
I guess other platforms are affected too.
We seem to have a known bug on some client NIC adapters that they send MLD report storms when they are powered off but Wake On LAN is enabled.
This causes lots of flooded MLD reports and this causes high (permanently about 80%) CPU load on our 2960 and 2960X switches (in interrupt mode in the SW Forwarding CPU Queue).
This happens even with default config and an IPv4 only SDM template.
Activating a IPv4/IPv6 SDM template and enable MLD snooping on the switch makes no difference.
debug platform cpu-queues software-fwd-q
Feb 12 2015 15:47:55.084 CET: SW-FWD-Q:IPv6 packet: Local Port Fwding L3If:Vlan1 L2If:FastEthernet0/1 DI:0x1F, LT:79, Vlan:1 SrcGPN:1, SrcGID:1, ACLLogIdx:0x0, MacDA:3333.ff9f.dcc2, MacSA: 901b.0e3f.c1a3 IPv6 SA:FE80::921B:EFF:FE3F:C1A3 DA:FF02::1:FF9F:DCC2 Nexthdr:0
TPFFD:D8000001_00008001_00A0005A-0000001F_FFFF0000_00000000
The lab switch runs 15.0(2)SE6.
There is no SVI on the 2960 for the VLAN where the MLD reports are received.
If I want to view the cause why they are punted to the CPU with "show platform forward", the switch crashes always.
switch#show platform forward f0/1 901b.0e3f.c1a3 3333.ff9f.dcc2 FE80::921B:EFF:FE3F:C1A3 FF02::1:FF9F:DCC2 icmp 131 0
