grg/parser-gen

Network packet parser generator

parser-gen's Issues

A good job in Programmable Data Plane! And can you help our group in the area?

I recently read your article “Forwarding Metamorphosis: Fast Programmable Match-Action Processing in Hardware for SDN" in SIGCOMM’13. In the paper, your team propose the RMT (reconfigurable match ta- bles) model, a new RISC-inspired pipelined architecture for switching chips. Great job! I am a graduate student of Xidian University, and my team is preparing to carry out a series of courses on "SDN" and "Programmable Data Plane" in our school. But because we don't have engineering experience in this area, we lack a RMT hardware engineering. Can you or your team help us build a simple hardware project in terms of RMT hardware, or recommend such an open source project for us?

C output?

I am curious if this tool (or some other tool) could take a packet description like this of a more complex protocol and turn it into a streaming parser for a C based daemon, in addition to hardware?

(my intent was to write a packet parser for the babel routing protocol, I was looking over peg, ran across your paper, and here I am)

I rather like the descriptive language....

AttributeError

There is an attribute error in this line:
from_fields = item.next_header.mapping.from_header.asList()

AttributeError: 'str' object has no attribute 'from_header'

headers-edge.txt mpls

I don't think you can reliably recognize protocol below MPLS (line 37..42)

You're now presuming reading BOS + IP version is reliable, but IANA assigns OUIs randomly since several years, so in many cases DMAC may look like IP version, causing misclassification in the naive approach.
Potentially causing reordering as ECMP guarantees might not apply.

I wonder if there even is proper/cheap fix? Or do you just need to assume IP based on IP version, until IP checksum fails (ipv4) => presume it's eth, or IPv6 until L4 checksum fails? Which seems terribly expensive.