On Thu, May 31, 2012 at 6:54 PM, Jesse Gross <je...@nicira.com> wrote: > On Thu, May 31, 2012 at 6:04 PM, ravi kerur <rke...@gmail.com> wrote: >> Thanks Jesse, I don't think I was clear. For mpls and qinq, offloading >> goes through ovs emulation code in vport-netdev.c. In addiition to >> TSO, there is checksum-offload and I believe this will be handled in >> the NIC and the driver sets the flags in DMA descriptor before sending >> to the NIC. This, I don't think will be handled in ovs offload code. >> Anyways, I will wait until you return and we can work on this. > > You can't assume that the NIC can do any offloading that it doesn't > advertise, including checksums. e1000e actually does advertise that > can handle generic TX checksum offloading (and apparently can to some > degree) but the driver doesn't handle it properly, which is a bug in > the driver. >
<rk> Agreed, I was trying to work-around this problem by un-setting skb->protocol. > The "offload" code in OVS is to not enable offloading in the NIC, it's > to provide a software emulation path in the cases where it can't be > handled in hardware (or other parts of the kernel). <rk> I understand this and yes I am using ovs offload emulation for mpls and qinq for the patches I sent today morning. I will look into this further from performance perspective. > > Your solution will not work for most NICs. If in doubt, you must emulate. <rk> agreed. _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
