After the definition section, we are prepared to start writing the body of the specification. This part of the specification shows how the bits in an instruction break down into opcodes, operands, immediate values, and the other pieces of an instruction. Then once this is figured out, the specification must also describe exactly how the processor would manipulate the data and operands if this particular instruction were executed. All of SLEIGH revolves around these two major tasks of disassembling and following semantics. It should come as no surprise then that the primary symbols defined and manipulated in the specification all have two key properties.
Formally a Specific Symbol is defined as an identifier associated with
The named registers that we defined earlier are the simplest examples of specific symbols (see Section 4.4, “Naming Registers”). The symbol identifier itself is the string that will get printed in disassembly and the varnode associated with the symbol is the one constructed by the define statement.
The other crucial part of the specification is how to map from the bits of a particular instruction to the specific symbols that apply. To this end we have the Family Symbol, which is defined as an identifier associated with a map from machine instructions to specific symbols.
The set of instruction encodings that map to a single specific symbol is called an instruction pattern and is described more fully in Section 7.4, “The Bit Pattern Section”. In most cases, this can be thought of as a mask on the bits of the instruction and a value that the remaining unmasked bits must match. At any rate, the family symbol identifier, when taken out of context, represents the entire collection of specific symbols involved in this map. But in the context of a specific instruction, the identifier represents the one specific symbol associated with the encoding of that instruction by the family symbol map.
Given these maps, the idea of the specification is to build up more and more complicated family symbols until we have a single root symbol. This gives us a single map from the bits of an instruction to the full disassembly of it and to the sequence of p-code instructions that simulate the instruction.
The symbol responsible for combining smaller family symbols is called a table, which is fully described in Section 7.8, “Tables”. Any table symbol can be used in the definition of other table symbols until the root symbol is fully described. The root symbol has the predefined identifier instruction.
Almost all identifiers live in the same global "scope". The global scope includes
All of the names in this scope must be unique. Each individual constructor (defined in Section 7, “Constructors”) defines a local scope for operand names. As with most languages, a local symbol with the same name as a global symbol hides the global symbol while that scope is in effect.
We list all of the symbols that are predefined by SLEIGH.
Table 2. Predefined Symbols
| Identifier | Meaning |
instruction |
The root instruction table. |
const |
Special address space for building constant varnodes. |
unique |
Address space for allocating temporary registers. |
inst_start |
Offset of the address of the current instruction. |
inst_next |
Offset of the address of the next instruction. |
inst_next2 |
Offset of the address of the instruction after the next instruction. |
epsilon |
A special identifier indicating an empty bit pattern. |
The most important of these to be aware of are inst_start and inst_next. These are family symbols which map in the context of particular instruction to the integer offset of either the address of the instruction or the address of the next instruction respectively. These are used in any relative branching situation. The inst_next2 is intended for conditional skip instruction situations. The remaining symbols are rarely used. The const and unique identifiers are address spaces. The epsilon identifier is inherited from SLED and is a specific symbol equivalent to the constant zero. The instruction identifier is the root instruction table.