Q: Where can I download OFP?
A: You can find instructions at the Download page.
Q: How can I join the OFP community?
A: Full instructions to join OFP can be found on this section.
Q: Are there any fees associated when using OFP?
A: No. The software is as-is, free of charge, licensed under a BSD license.
Q: What protocols are supported in OFP?
A: it Supports IPv4 and IPv6, TCP, UDP, IGMP and VXLAN (do we support UDP, TCP and IGMP v6 variants?)
Q: Does OFP always need ODP to work?
A: Currently the stack interfaces to the ODP API only. This is for maximum portability which has been proven in the early cross platform validation.
Q: Does OFP work with DPDK?
A: Yes. Support for DPDK can be made with ODP-DPDK integration layer from Linaro. ODP 1.4 is currently validated with DPDK 2.1 and is maintained as DPDK progresses. Check in with the project at any given time for the latest compatibility.
Q: What Linux kernel versions are supported?
A: OFP is not dependent on kernel versions
Q: Is OFP launched as a product?
A: No. OFP has been developed as an incubation project and now launched to the open source community. We are inviting open source community to contribute to the program.
Q: When will OFP be product quality?
A: It is premature to speculate on this when we are just launching from the incubation phase. OFP board together with the contributors will decide when OFP has reached production maturity
Q: What is the performance target for OFP?
A: OFP is designed to be the fastest IP stack for the Linux user space.
Q: What processor architectures does OFP run on?
A: OFP is processor ISA agnostic and has already been validated on ARM, MIPS and x86 architectures.
Q: Does the OFP IP stack mimic Linux stack config inside the fastpath, meaning does it intercept Linux ipconfig/ip commands and automatically create similar entities inside the fastpath stack?
A: Yes. Uses Netlink to sync with ifconfig commands and also with routes.
Q: Does the OFP IP stack have full IPv4/IPv6 fragmentation/re-assembly support?
A: IPv4 only. Fragmentation and reassembly on IPv4.
Q: Does the OFP IP stack have full TCP IPv4/IPv6 TCP segmentation support?
A: Yes. The TCP application payload sent will be segmented into pieces as determined by the maximum segment size (MSS). TCP fragmentation and MSS implementation are ported from libuinet (NetBSD). We don’t support TCP segmentation offload (TSO).
Q: Does the OFP IP stack have a replacement/equivalent of DPDK KNI (Kernel Network Interface) to allow forwarding packets between Linux/Fastpath?
A: OFP uses a generic solution called TAP, not intended for performance. Fastpath can relay packets to Linux(slowpath). The use case to relay packets from Linux to Fastpath is not implemented. Linux will send packets straight to NIC/wire.
Q: Does the OFP IP stack have VLAN/VXLAN support?
Q: Does the OFP IP stack have support for IPv4/IPv6 routing?
A: Yes. OFP supports TCP and IGMP which most routing protocols use
Q: Does the OFP IP stack have ip source routing and multiple routing tables support?
A: Not for IP source routing, but yes, for “Multiple routing tables”, one can use VRF support.
Q: Does the OFP IP stack have any ability to learn host Linux iptables and accelerate these iptables/ip6tables within the fastpath?
A: No, no learning of iptables rules implemented.
Q: What versions of ODP does OFP support?
A: At the moment it supports all the way from 1.1.0 to 1.4.0. ODP 1.5.0 patches are in.