Register Access Implementation Guide

A ASSP hardware register is a physical interface to an item of +hardware which contains values, such as bit fields, for controlling +ASSP features.

Pre-requisites

Before the implementing a register access implementation, the +following must be decided:

The area in memory to which the registers are mapped. The start +of this memory region is known as the base address.
The location of each register in terms of an offset from the +base address.

The above information will be present in a specific header +file which is used by any device driver that uses this API. This header +file also contains other information required to access the registers, +such as:

constants for bit field offsets
bit field values

Implementation

To provide register access, hardware implementers need to implement +the AsspRegister class.

Generic access to a register is provided in the form of three +operations on a hardware register: read, write and modify. All these +functions can be called in any context.

Symbian platform provides +the class header and inline implementations of the 8, 16 and 32-bit +wide functions. You must implement all the register modify functions +and all functions to access 64-bit wide registers. You may also need +to replace the inline 8, 16 and 32-bit wide register read and write +functions if the ASSP contains write-only registers.

You must +implement these functions atomically so that there is no possibility +of another thread or interrupt service routine accessing the contents +of the register before the function returns. Where you cannot implement +the function with a single machine instruction you must disable interrupts +during the register operation (if you are working with a single core +architecture) or wrap a spinlock around the register operation (if +you are working with a multicore architecture).

There are +three cases in which you must enforce atomicity with interrupt disabled +or use of spinlocks.

The modify operations +involve reading the register, taking a local copy, changing the local +copy and writing it back to the register and cannot be performed with +a single machine instruction.
Operations on +write-only registers involve maintaining a copy of the current contents +of the register within the AsspRegister class. All write and modify operations therefore cannot be performed +with a single machine instruction.
Some architectures +contain 64-bit registers. In this case, 64-bit operations involve +at least two machine instructions.

Baseport implementations of the Register Access platform service +come in the form of a ASSP library (or DLL). To avoid the need for +each device driver which uses the ASSP register class having to specify +a baseport specific library file corresponding to the ASSP library, +the name of the baseport specific library file should instead be included +in a file used to define other variant specific settings such as the +CPU core type, the memory model and the name of the variant. Each +baseport will have a file called variant.mmh, +where variant is the name of the baseport to be used. +An example of this being used in a variant.mmh file is:

#if defined(__USING_BASPORT__) +library VariantTarget(basport_name,assp_library) +#endif +baseport_name is the name of the variant and assp_library +is the name of the ASSP specific library file.

In order to +specify which baseport assp register set up is to be used in the client +projects for this baseport, the following line is included in the +MMP file of those projects

#define __USING_BASEPORT___