Subrata Banik - System Firmware: An Essential Guide to Open Source and Embedded Solutions

This Apress imprint is published by the registered company APress Media, LLC, part of Springer Nature.

The registered company address is: 1 New York Plaza, New York, NY 10004, U.S.A.

This book contains key aspects of an evolution that is about to happen for system firmware (a.k.a. the BIOS). The last similar scale change in system firmware happened in the late 90s, from BIOS to UEFI. While the birth of the BIOS (e.g., the separation of system firmware from the operating system) in the 70s led to the era of the PC, UEFI laid a solid foundation for the success of computer desktops and servers, which is the foundation of the modern Internet world.

Now, with the Internet moving toward cloud services and artificial intelligence, will the innovation of UEFI continue to meet needs or will there be an evolution in system firmware?

Thomas Khun described in his book The Structure of Science Revolutions the concepts of paradigms and revolutions. These concepts apply not only to scientific research but also to the computer industry. We normally operate in a paradigm. A well-established paradigm is very useful and powerful, and thus we tend to neglect stuff outside of the paradigm. As time goes by, the anomalies pile up and then a paradigm shift happens, which is called a revolution.

One anomaly is the shift to cloud computing. Hyperscalers need to manage servers at scale, and in the meantime they often have feature differentiation and shortened time to production (TTP) needs.

OSF allows hyperscalers to own the system firmware more easily, since a smaller engineering team will be capable of developing and managing it at scale, enabled by industry collaboration. With Linux playing an instrumental role in OSF, Linux engineers are turned into system firmware engineers, therefore system firmware can be managed at scale just like Linux. In addition, new feature support and bug fixes in Linux can be leveraged in system firmware directly, leading to improved stability and reduced security vulnerability.

Another anomaly is the arrival of the computer architecture golden age. David Patterson declared that, like the 1980s, the next decade will be exciting for computer architects in academia and in industries. Not only scientific research but also societal advancements (such as Web 3.0 and the metaverse) demand exponential growth in computational power. On top of that, the focus on computation is shifting from general purpose computing to artificial intelligence workloads. However, Moores Law is not supported by chip process technological improvements alone any longer. All of these changes require a new generation of server processors and server systems. New technologies and design philosophies beyond the von Neumann architecture have been springing up. Other examples include heterogeneous processor design enabled by chiplet and UCIe, and new interconnect technologies such as CXL, to name a few. All of these new technologies require timely development of system firmware as part of holistic solutions.

In the meantime, another event that happened in the industry was the maturation of Linux into the de-facto standard of the OS industry. Linux answers computer technology advancements gracefully and manages versatility effectively. Why should we develop software for the same technology twiceonce for Linux, and another time for system firmware? Why should we stabilize and harden software for the same technology twice? Why should we require two vastly different skill sets for system firmware engineers and kernel/OS engineers?

You might ask whether OSF is mature enough for prime time. This is an important question. After all, technology alone is not good enough; we see too many cases where a good technology does not gain traction in the industry.

With over 10 years of technology accumulation, FSP (Firmware Support Package) technology, as Intels silicon initialization binary, matured. FSP is a Plan-Of-Record for Intel server processors, starting with Intels Sapphire Rapids Scalable Processor. In addition, as part of Intels OneAPI initiative, Vincent Zimmer coined Universal Scalable Firmware (USF) architecture, which supports OSF.

On the other hand, coreboot has been shipped in hundreds of millions of devices, notably chromebook, among other products. LinuxBoot has been running at-scale at some of the worlds largest data centers.

Those components together, as open system firmware for servers, have seen great progress in the last several years. In 2021, OCP (Open Compute Project) accepted OSF for OCP DeltaLake server (based on Intel Copper Lake Scalable Processor) and is available for the community. An OCP community lab is being built to enable access to OCP gears, with DeltaLake servers included, for the community. OSF is planned to be deployed in production in ByteDances data centers.