Common questions about UEFI include:
How are programs in UEFI implemented?
What makes UEFI programming different from an operating system?
What makes UEFI different from other firmware environments?
In particular, what is the programming model for a UEFI Driver?
Key points about writing UEFI-conformant drivers are that:
UEFI Drivers are relocatable PE/COFF images whose format is defined by the Microsoft Portable Executable and Common Object File Format Specification.
UEFI Drivers may be compiled for any of the CPU architectures supported by the UEFI Specification.
UEFI Drivers run on a single CPU thread.
The driver support infrastructure does not extend beyond the boot processor.
Drivers sit above some interfaces (for example, bus abstractions) and below other interfaces: They are both consumers and producers. The UEFI Specification defines the interfaces and they are extensible.
Each driver is expected to cooperate with other drivers, other modules and the underlying core services.
The communicating modules bind together to create stacks of cooperating drivers to accomplish tasks.
Inter-module communication is enabled via interfaces known as protocols and via events.
Tables provided at invocation provide access to core services.
The operating environment is non-preemptive and polled. There are no tasks per se. Instead, modules execute sequentially.
There is only one interrupt: the timer. This means that data structures accessed by both in-line code and timer-driven code must take care to synchronize access to critical paths. This is accomplished via privilege levels.