TABLE OF CONTENTS While some packet-processing applications are more complicated than others, if you think about most network processor applications,the basic receive-process-transmit framework, as shown in Figure 3.1, most likely applies. Whether you are writing code for a switch that simply forwards packets, or you are building a content-aware load balancer that decrypts packet contents and performs string searches of URLs, your application receives, processes, and then transmits packets. The only difference lies in the complexity of the processing tasks.
Given this, a good place for you to start is with the most basic receive-processing-transmit application: counting packets. This approach shows you how to receive packets, get them to a place where they can be processed, and finally get them from there to the transmit task. Following this, you can extend the processing task to include bigger and better things than just counting packets.
To simplify the code in this chapter, all of the receive, process, and transmit functions execute in a single thread in their respective microengines. This way, synchronization methods on the microengines and complicated data structures and algorithms normally associated with a meaningful processing task are not needed. After all, this chapter contains the first serious piece of code in the book! Subsequent chapters deal with all of the issues avoided in this chapter.
A packet processing application without packets is fairly boring. So the first step to building an interesting application is to receive packets. Once a packet is received, it can be passed to the...
