include "dis.m";
dis := load Dis Dis->PATH;

Inst: adt
{
	op:		int;
	addr:	int;
	mid:		int;
	src:		int;
	dst:		int;
};

Type: adt
{
	size:	int;
	map:		array of byte;
};

Data: adt
{
	op:	int;		# encoded op
	n:	int;		# number of elements
	off:	int;		# byte offset in data space
	pick {
	Zero =>		# DEFZ
	Bytes =>		# DEFB
		bytes:	array of byte;
	Words =>		# DEFW
		words:	array of int;
	String =>		# DEFS
		str:	string;
	Reals =>		# DEFF
		reals:	array of real;
	Array =>		# DEFA
		typex:	int;
		length:	int;
	Aindex =>		# DIND
		index:	int;
	Arestore =>	# DAPOP
	Bigs =>		# DEFL
		bigs:		array of big;
	}
};

Link: adt
{
	pc:		int;
	desc:	int;
	sig:		int;
	name:	string;
};

Import: adt
{
	sig:	int;
	name:	string;
};

Except: adt
{
	s:	string;
	pc:	int;
};

Handler: adt
{
	pc1:	int;
	pc2:	int;
	eoff:	int;
	ne:	int;
	t:	ref Type;
	etab:	array of ref Except;
};

Mod: adt
{
	name:	string;
	srcpath:	string;

	magic:	int;
	rt:		int;
	ssize:	int;
	isize:	int;
	dsize:	int;
	tsize:	int;
	lsize:	int;
	entry:	int;
	entryt:	int;

	inst:	array of ref Inst;
	types:	array of ref Type;
	data:	list of ref Data;
	links:	array of ref Link;
	imports:	array of array of ref Import;
	handlers:	array of ref Handler;

	sign:	array of byte;
};

init:	fn();
loadobj:	fn(file: string): (ref Mod, string);
op2s:	fn(op: int): string;
inst2s:	fn(i: ref Inst): string;

Init must be called before any other function, to initialise the module.

Loadobj reads a Dis object file from file, and returns a reference to a Mod adt that represents the module's contents, as the first element of the tuple; the string element of the tuple is nil. On error, the string element contains a diagnostic, and the reference is nil.

Op2s returns the assembly-language representation, as used by asm(1), of the Dis operation code op. It returns the string `OPop' if op does not correspond to a known operation code.

Inst2s returns a string corresponding to a disassembly of Dis instruction i, including addressing modes.

The module defines integer constants giving symbolic names to the Dis instruction codes, all of the form Iname where name is the name of the instruction, all in upper case:

INOP, IALT, INBALT, ... INEWZ, INEWAZ, IRAISE

The name MAXDIS is also defined; it has the value of the first unassigned Dis operation code.

Most of the members of the adt types have an obvious interpretation on reference to dis(6).

The adt Mod represents a single module. It contains values extracted from the module's header, and references to structures representing the contents of the Dis file's code, data, type and external linkage sections:

magic

The constant XMAGIC (unsigned Dis module) or the constant SMAGIC (signed Dis module).

sign

If magic is SMAGIC, the sign field contains the bytes in the signature section of the module header. Otherwise, there is no signature and sign is nil.

name

The name of the implementation module.

srcpath

The source of the dis file relative to the inferno root.

rt

Run-time options: a bit mask of the constants MUSTCOMPILE, DONTCOMPILE and SHAREMP.

ssize

Stack extent

isize

Number of instructions

dsize

Size in bytes of the module's global data area

tsize

Number of type descriptors

lsize

Number of external linkage descriptors

entry

PC (instruction offset) of the default entry point for the module

entryt

Index of the type descriptor for the module's entry point

inst

Array representing the contents of the code segment; length m.isize

types

Array of the module's type descriptors; length m.tsize

data

list of data descriptors representing instructions for creating the module's data segment

links

array of the module's external linkage descriptors (for exported functions); length m.lsize

imports

an array of import descriptor tables, one table for each module imported by this module. Each table is an array of pairs giving the signature and name of each function imported.

handlers

an array of exception handlers used in this module. Each handler consists of the range of pc's it covers, the exception structure offset within the frame, the number of declared exceptions (as opposed to strings) in the handler, the type (if any) of any memory to clear when the exception occurs and a table of exceptions. The latter is an array containing pairs of exceptions and pc values. The final entry gives the pc to jump to in the '*' case or -1 if not applicable.

The Type adt represents the value of a type descriptor:

size

Size in bytes of the object represented by this descriptor

map

Bitmap describing the location of pointers in the object (see dis(6))

The Link adt represents the value of a link descriptor:

name

Name of the exported function

pc

Instruction index in Mod.code of the function's entry point

desc

Index in Mod.types of the type describing the function's stack frame

sig

Integer hash of the function's type signature

The Inst adt represents a single Dis instruction in the instruction stream. The member op is the Dis instruction code. The member addr contains the addressing mode flags for middle, source and destination operands. Constants are defined to help unpack it.

The middle operand description is selected by the constant mask ARM:

i.addr & ARM

The valid results and interpretation are as follows:

AXNON

No middle operand.

AXIMM

$n

AXINF

n(fp)

AXINM

n(mp)

The source operand's addressing mode is extracted as follows:

(i.addr>>3)&AMASK

The following combinations are valid, where n is the value in i.src:

AXXX

No operand

AFP

The operand is n(fp)

AMP

The operand is n(mp)

AIMM

The operand is $n (ie, immediate literal n)

AIND|AFP

The operand is si(fi(fp))

AIND|AMP

The operand is si(fi(mp))

where fi is the offset for the first indirection, extracted from n:

(n>>16)&16rFFFF),

and si is the offset for the second indirection, also extracted from n:

(n&16rFFFF).

The destination addressing mode is interpreted in a similar way, except that the addressing mode is extracted as follows:

(i.addr&AMASK)

and the value of the offset n is found in i.dst. Fi and si are extracted from n as before.

Finally, Data adt represents a data item, which tells the system's module loader how to initialise part of the module's global data segment. It has the following members:

op

the encoded type and length; usually ignored: the pick tag and n, below, usually suffice

n

the number of data values

off

the byte offset of the first data value to initialise, relative to the current loading base

The alternatives of the pick select the correct variant to see the data values encoded in the object file as Limbo values of the correct type. The interpretation is straightforward for the tags Bytes, Words, Bigs and Reals: the corresponding array members are arrays of n elements of the appropriate type. The remaining cases are as follows:

String

The member str has the decoded representation of the corresponding n data bytes from the object file.

Array

The member typex is the index in Mod.types of the array's type, and member length is its length.

Aindex

This alternative can appear only following a value of Data.Array. The member index is an index into the corresponding array as represented in the global data space, which determines a new loading base address for subsequent Data items. The previous base address is stacked on an internal stack.

Arestore

Pop the address from the internal address stack and make that the current loading address. The request marks the end of a sequence of Data items initialising an array.