VRF MTU issue

[sidog]

I am having some kernel crash issues when using VRF that are occurring during transmit fragmentation.

I am running Linux 4.8.3 on a x86_64 CPU.

Information on the VRF follows:

# ip -d link show type vrf mgmt
168: mgmt: <NOARP,MASTER,UP,LOWER_UP> mtu 65536 qdisc noqueue state UP mode DEFAULT group default qlen 1000
link/ether 42:56:a7:71:b7:cc brd ff:ff:ff:ff:ff:ff promiscuity 0
vrf table 31 addrgenmode eui64 numtxqueues 1 numrxqueues 1

# ip link show vrf mgmt
132: craft1.2: <BROADCAST,MULTICAST,PROMISC,UP,LOWER_UP> mtu 1500 qdisc noqueue master mgmt state UP mode DEFAULT group default qlen 1000
link/ether cc:be:59:54:45:2e brd ff:ff:ff:ff:ff:ff
137: system-craft: <BROADCAST,MULTICAST,PROMISC,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast master mgmt state UNKNOWN mode DEFAULT group default
link/ether cc:be:59:54:29:2e brd ff:ff:ff:ff:ff:ff

# brctl show craft1.2
bridge name bridge id STP enabled interfaces
craft1.2 8000.ccbe5954452e no eth1

# ip link show eth1
4: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq master craft1.2 state UP mode DEFAULT group default qlen 1000
link/ether cc:be:59:54:45:2e brd ff:ff:ff:ff:ff:ff


If I set the MTU of the VRF (mgmt) to 1500, the issue is resolved and no more kernel crashes occur.

Are there any MTU restrictions for a VRF in relationship to enslaved devices?

I have been unable to find any information on this subject so far.

Thanks

 

[Lazydog]

There should not be as long at the remote system is capable of receiving large packets. What are you setting this to when the kernel crashes?

 

[sidog]

I think the server MTU is 9k. The VRF (mgmt) MTU is 65536 the enslaved bridge (craft1.2) MTU is 1500 and the ethernet port (eth1) that is in the bridge has an MTU of 1500.

 

[Lazydog]

OK here the thing you have to remember in order to get full Jumbo packets from point to point everything along the path must support jumbo frames. If just one device doesn't you'll never get jumbo frames working.